Author: SK

Move to Substack?

Ever since I acquired this domain name (back in 2008), this blog has been hosted on WordPress. However, now I’m starting to wonder if I should port it to Substack (the URL will remain the same). Let me explain why.

Recently, the admins at Substack wrote an article that “blogging boom is back“. Quoting,

What we’re seeing now feels a lot like that early blogging boom. There was an intimacy we felt reading our favorite blogs, a personal connection to the writers and the communities that grew around them. We stacked our Google Reader with their RSS feeds and turned to them for restaurant recommendations, recipes, home decor trends, crafting inspiration, gossip, political analysis, and life advice. Writers on Substack are providing that same intimacy and connection with the communities they create. No media conglomerates edit their words and ideas. We have access to our favorite writers, just as we did in those fast blogging days. We see ourselves in the personal stories they share; we trust them.

Having started (and not really continued) some five or six substacks in the last five years, I broadly agree to this sentiment. And the reason why I feel like Substack might be “the old blogging in a new form” is due to comments. People actually comment on Substack, unlike on WordPress.

I was reading a friend’s substack yesterday, and noticed a possibly Freudian slip. I quickly hit the comment button and shot off a comment. As of this morning, he has already seen it and replied. This was a regular feature on the blogosphere in the late 2000s. In the 2010s, this kind of behaviour simply died.

And coming to think of it, it possibly has to do with user experience. The user experience for commenting on WordPress absolutely sucks. There is no concept of login across sites (I don’t know when this “openID” was last maintained). So every time you need to write a comment, you need to enter your name, email, website, and maybe even fill a captcha before your comment gets accepted. And that is IF your comment gets accepted (the number of times I’ve written an elaborate comment and seen it NOT go through is insane).

Two decades ago, when blogging was big and growing rapidly, people used to leave comments on one another’s blogs. A LOT. In fact, I discovered a lot of blogs I followed back in the day by following comment trails from blogs that I already liked. It could occasionally be riotous, like on this legendary post by Ravikiran Rao.

And then sometime in 2010 or 11, it all stopped. Almost all of a sudden. I have statistics on number of comments per post on this blog over the years, but right now NED to pull it up (and also, the commenting was heavier when this site was on LJ). Suddenly people stopped commenting. And I think this had an impact on blogging as well. With little or no feedback, people didn’t feel like writing.

People started talking about the death of blogs. Sometimes, writing this would feel like shouting into an empty room. Then again, I’ve considered this as a documentation of my life and my thoughts (and your benefits, if any, being strictly collateral), and carried on.

What Substack seems to have shown is that the appetite for blogging didn’t go away. The appetite for commenting also didn’t go away. It’s only the user experience. Over the years, maybe coinciding with WordPress being the dominant blogging platform (and WordPress being more popular for making websites than for blogs), the user experience of commenting deteriorated. And as people commented less, they blogged less.

Now, looking at the comment density on Substack, I’m seriously considering if I should make a shift there. Still need to see how easily I can port all of this stuff without breaking. But if I can, I might just. What do you think? Do leave a comment (and if you think this blogpost is too hard to comment on, maybe comment on this “note” instead).

Brahmastra

Sometimes we overdo “option value”. We do things that have a small possibility of a big upside, and big possibility of no or very minimal downside, in the belief that “nothing can go wrong in trying”.

My father used to term this “pulling a mountain with a string”, with the reasoning being that if you actually manage to pull, then you have moved a mountain. If not, all that you have lost is a string.

There is one kind of situation, however, where I think we might overindex on option value – these are what I call “one shot events” or “brahmastras”.

Going into a little bit of mythology, there is the story of the Brahmastra in the Mahabharata. Famously, Karna possesses it. It is an incredibly powerful weapon with the feature (or bug, rather) that it can be used only once. Karna would have set it aside to use on Arjuna, but the Pandavas decide to send Ghatotkacha to create havoc during the night fight when Karna is forced to use up his brahmastra on Ghatotkacha – meaning he didn’t have access to it in his battle with Arjuna, where he (Karna) ultimately got killed.

Because the Brahmastra could be used only once, Karna wanted to maximise the impact of the weapon. His initial plan was to use it on what he thought might be a decisive battle with Arjuna. The Pandavas’ counterplan was to force him to use it earlier.

Actually, thinking about it – the Brahmastra can be thought of as another kind of option. The problem here being one of optimal exercise. Actually, there is a very stud paper written by economist Avinash Dixit on this topic – regarding Elaine’s sponges.

Read the whole paper. It is surely worth it. To quickly summarise, Elaine has a limited number of “contraceptive sponges”, and wants to maximise her “utility” of using them. When a guy comes along, she needs to decide whether it is worth expending a sponge on him. Dixit derives a nice equation to determine a function for this.

Basically, Brahmastra occurs when you have only one sponge left, and you need to use it at an “optimal time”. There is another problem in economics  called the “secretary problem” (nothing to do with secretary birds) that deals with this.

Recently I’ve been thinking – these kind of Brahmastra / sponge / secretary problems are important to solve when you are thinking of talking to someone.

Let’s say you have what you think is a studmax application of GenAI and want to talk to VCs about it. If you go too early, the VC will only see a half-baked version of your idea, and even if you go to them later once you have fully formed it, the half-baked idea you had showed them will influence them enough to discount your later fully formed idea.

And if you go too late, the idea may not be that studmax any more, and the VC might dismiss it. So it’s a problem of “optimal exercise” (note that this is an issue only with American options, not European).

It is similar with asking someone out (or so I think – I’ve been out of this business for 14 years now). You approach them “too early” (before they know you), they will dismiss you then and then forever. You approach too late and the option would have expired.

In the world of finance, we focus too much on the PRICE of options and (based on my now limited knowledge) too little on optimal expiry of the said options. In the real world, the latter is also important.

Mid Wife Crisis

A month or two ago, my wife wrote this studmax blogpost about “Mid Life Crises“. I read it when I was in draft. I read it again soon after it was published. And I read it again, and again, and again. And I’ve been showing it to a lot of people I meet.

Having seen this post, it has been impossible to unsee. I used to wonder why Hospi puts so many dance videos on Instagram, and then realised it must be mid-life crisis. I saw Josh go from being a VLSI Engineer to pushing limits, and have classified that as mid-life crisis as well. Recently, Ruddra posted topless gym photos. I actually commented that it “must be mid life crisis”.

A few days ago, my wife and I were talking about an acquaintance who is possibly having an affair, and that again we put down to “mid life crisis”. When asked why he had had a second child eight years after his first, a relative had said “early forties you hit mid life crisis. We had the option of a sports car or another kid. Decided to go for the latter”.

You see – once you’ve seen this concept, and if you belong to the broad 35-50 age group, it is absolutely impossible to un-see. You get excited about possibly meeting an old flame after ages, and mid-life crisis explains that. You want to get hammered when you meet your old friends, and mid-life crisis explains that as well.

Now, being 40 and having identified that I’m at “mid life”, the challenge is to deal with the crisis in the least destructive manner possible. That the crisis exists, and needs to be dealt with, is a fact. The choice is in terms of how to deal with it.

If you are married (or in a long term relationship), you don’t want to have an affair and put that in jeopardy (also the question needs to be asked if one’s demand for an affair can be met with supply). Alcohol and dope can be destructive to your physical and mental health. Social media gives you dopamine (and I think that was my main tool of dealing with mid-life crisis till recently), but can be both addictive and anxiety-causing. Gaming (I’ve considered buying a console for nearly 2 years now, but I’m yet to act on it) gives you the dopamine but again comes with addiction risk.

And that is where the “healthy obsessions come in”. You notice that many people around my age take to running, or cycling, or lifting, or a combination of these. All of these are physically strenuous activities, and thus can give you your endorphins. As an added bonus, they increase either your cardiovascular health or your muscle strength, thus preparing you better for old age.

So you find these to be fairly common midlife-dealing hobbies. If you can find one of these that you don’t get bored by, it’s a clear win-win (I lift weights). The “problem” is that because it is physically strenuous you can only do so much of it (I go to the gym for an hour each 2-3 times a week. If you do less strenuous stuff like running, you can perhaps do an hour 7 times a week). And so you need more. The question is what you can choose that is non-destructive (relatively speaking).

A month or two back, I got introduced to online quizzing. I’m currently playing three leagues (one of them will end soon, but I’ve signed up for yet another starting at the end of this month). I find that this is a fairly high-dopamine hobby (to the extent that now I’ve started doing this late in the night since it interferes with my sleep), and keeps me stimulated enough. The only issue is it can be potentially addictive (when I played my first “friendly” online quiz 2-3 weekends ago, I felt the same way as I did when I smoked my first cigarette – “this is addictive shit, I need to be careful”).

Interestingly, two friends I spoke to in the week of August 15th (and I’m pretty sure those two have never met or spoken to each other) told me “start working on another book to deal with your mid life crisis”. I admit they too have a point, but I don’t know what to write about.

But yes – midlife crisis is a real thing. It is only about how you choose to deal with it without causing self-harm.

SHAP and WAR

A few months back, at work, a couple of kids in my team taught me this concept called “SHAP“. I won’t go into the technical details here (or maybe I will later on in this post), but it is basically an algo that helps us explain a machine learning model.

It was one of those concepts that I found absolutely mind-blowing, to the extent that after these guys taught this concept to me, it became the proverbial hammer, and I started looking for “nails” all around the company. I’m pretty sure I’ve abused it (SHAP I mean).

Most of the documentation of SHAP is not very good, as you might expect about something that is very deeply technical. So maybe I’ll give a brief intro here. Or maybe not – it’s been a few months since I started using and abusing it, and so I’ve forgotten the maths.

In any case, this is one of those concepts that made me incredibly happy on the day I learnt about it. Basically, to put it “in brief”, what you essentially do is to zero out an explanatory variable, and see what the model predicts with the rest of the variables. The difference between this and the actual model output, approximately speaking, is the contribution of this explanatory variable to this particular prediction.

The beauty of SHAP is that you can calculate the value for hundreds of explanatory variables and millions of observations in fairly quick time. And that’s what’s led me to use and abuse it.

In any case, I was reading something about American sport recently, and I realised that SHAP is almost exactly identical (in concept, though not in maths) to Wins Above Replacement.

WAR works the same way – a player is replaced by a hypothetical “average similar player” (the replacement), and the model calculates how much the team would have won in that case. A player’s WAR is thus the difference between the “actuals” (what the team has actually won) and the hypothetical if this particular player had been replaced by the average replacement.

This, if you think about it, is exactly similar to zeroing out the idiosyncrasies of a particular player. So – let’s say you had a machine learning model where you had to predict wins based on certain sets of features of each player (think of the features they put on those otherwise horrible spider charts when comparing footballers).

You build this model. And then to find out the contribution of a particular player, you get rid of all of this person’s features (or replace it with “average” for all data points). And then look at the prediction and how different it is from the “actual prediction”. Depending on how you look at it, it can either be SHAP or WAR.

In other words, the two concepts are pretty much exactly the same!

LLMs and Software Margins

A few months back, I came across this article that talked about margins in the software industry. Long ago, computer software was well known to be an insanely high gross margin industry. However, it is not the case any more.

If you look at SaaS (software as a service) companies, a lot of them barely make much profits any more. So what changed?

The answer is infrastructure. In the olden days, when all hardware was “on premise”, software would be a bunch of lines of code that would get sold, and then run on the client’s on-premise hardware. Thus, once the code had been written and tested and perfected, the only cost that the vendor faced was to install the code on the client’s hardware (including the cost of engineers involved in the installation). And the margins soared.

Then (I’m still paraphrasing the article that I had read, and now can’t find), the cloud happened. Hardware wasn’t all on-premise any more. People figured out that software could be sold “as a service” (hence SaaS). Which means, instead of charging for installing some code on a computer, you could charge for API hits, or function calls. Everything became smooth.

The catch, though, was that the software would now have to be hosted on hardware maintained (in the cloud) by the vendor. Which meant now the marginal cost of delivery suddenly became non-zero. Rather, it went from O(1) (one time installation) to O(n) (costing each time it gets hit, or the time for which it is maintained). And this had a material impact on software margins.

I’m thinking of this now in the wake of new-fangled open source LLMs that keep getting announced every day. Every new LLM that comes out gets compared with ChatGPT, and people tell you that this new LLM is “open source”. And you get excited that you can get for free what you would have to pay for with ChatGPT.

Of course, the catch here is that ChatGPT is like SaaS – not only does it provide you the “LLM service” it also hosts the service for you and answers your questions, for a fee.

These open source models are like the traditional “on-premise” computer software industry – they have good code but the issue of course is that you need to supply your own hardware. Add in the cost of maintaining the said hardware, and you see where you might spend with the open source LLMs.

That said, Free != Open Source. The Open Source LLMs are not only free, but also open source – and so, the real value in them is that you can actually build on the existing algorithms and not have to pay a fee (except for your own infrastructure).

And from that perspective, it’s exciting that so many new tools are coming along.

Party Games

A year and half back, my wife had gone to Gurgaon on work. One evening, she called and told me that she was “going to go for a party at the guest house”, which I duly conveyed to our daughter.

The next morning, our daughter woke up and asked me about her mother’s party. Having been appraised of the proceedings late in the night, I shared the summary. “That is all fine, Appa”, she want, “but WHAT WAS THERE at the party?”.

I was a bit puzzled by the question and said there was nothing. “Why does a party need to have anything?”, I replied, “in this case there was big people juice, which people drank and talked to each other”.

It was in the course of that conversation that I realised that most kids’ parties usually have “something”. Some have bouncy castles. Some take place in play areas. Some people organise magic shows. Others have art workshops. And so on. A lot of kids’ parties are “structured”, with “stuff to do”.

Coming to think of it, this is not true of kids’ parties alone. Even a lot of adult parties nowadays have “themes”. So people have “poker nights”, or “board game nights”, or “movie nights” for which they call other people and socialise and together perform what can sometimes be a perfectly satisfactory single player activity.

Poker nights, I can understand, since it is sport, and one that can be much better played offline. However, I can’t imagine calling a bunch of random friends for a “poker night” – if it’s a poker night, it ought to be a bunch of people who are also interested in poker.

That aside, why should you bother hosting a party for a bunch of friends, and then not give them the opportunity to talk to one another, and instead subject them to some “party game”? “What is even the point of having structured activities at a party?”, my wife wondered loudly one morning.

My theory is this – not everyone is interesting and capable of holding an intelligent conversation. However, everyone has the need to talk to other people and socialise.

So if you are not sure about the quality of conversations that the people you are inviting to a party are likely to contribute, you want to somehow ensure that the party is at least somewhat interesting to everyone that attends. And so, you get rid of the upside (of some fantastic conversation happening at the party), and instead limit the downside (of everyone there getting bored), and put some structured activity on the party.

In other words, you put a “collar” on the party.

Collar – a derivatives strategy where you give up on upside to avoid downside

I have written here about the concept of “alcohol buddies“:

My friend Hari The Kid has this concept of “alcohol buddies”. These are basically people who you can hang out with only if at least one of you is drunk (there are some extreme cases who are so difficult to hang out with that the only way to do it is for BOTH of you to be drunk). The idea is that if both of you are sober there is nothing really to talk about and you will easily get bored. But hey, these are your friends so you need to hang out with them, and the easiest way of doing so is to convert them into alcohol buddies.

Now, there are some people who you can’t hang out with in “ground state”, but when one or both of you is drunk you can have an interesting conversation. Those are alcohol buddies.

However, there is a (possibly small) set of people who are fundamentally so uninteresting that even if both of you are pissed drunk, it is impossible to have a conversation is interesting to both people. And if you are having a largish party with a diverse set of guests, it is likely that there are many such pairs of guests, who cannot talk to each other even when pissed drunk.

And that is where having a party game helps. It prevents people from having random conversations and instead corrals (notice that wordplay there) everyone into the party game collar. No upside, no downside, nobody needs to find that there are others at the party who are absolutely boring to them. They all go home happy.

So far, we have resisted this “themed party” concept, except maybe in the context of NED Talks. Even our daughter’s birthday parties, so far, have been at home (once in Lalbagh during the pandemic), with the only “planned activity” being eating cake and snacks, and kids randomly playing in her room.

Let’s see how far we can carry this on!

Back to IIT

I hereby regret to inform you all that Sri Gurunath Patisserie, and everything around it including the Gurunath Stores and Moon Lab, is no more. There is no sign of its existence. Some new building, maybe an extension to the “giga mess” next door is coming up there.

Until this part of my walk early in the morning on the 15th of August, I had been thrilled to be back at IITM. Thanks to the kindness o the IITM Alumni Association, I had managed to get a room at the Taramani Guest House there for the morning, after my train had deposited me in Chennai at 4:15 am.

Of all the “institutes of national importance” I’ve visited in the recent past (last 1-2 years), IITM seemed the most friendly (and IIMB the least). This is ironic since as a student it was absolutely the other way round – the administration in IITM didn’t treat students well at all but in IIMB they were rather nice. However, now, post-pandemic IIMB has become a terror to get into, with some insane protocols and rituals.

At the IITM gate, though, all I had to do was to show a mail that I had a booking in Taramani House, and my auto was waved past. Delhi Avenue looks exactly the same as it did two decades ago, though maybe some new faculty blocks have been added to the sides. The stadium now has a sponsor (some Watsa – no Prem), and Gajendra Circle was all lit up for Independence Day. Else it was a very familiar ride in.

Taramani House has turned around, though, with its entrance now facing the road between CLT and OAT, and not towards Narmada (my old hostel). The auto dropped me there, and I duly handed over the ?160 change I’d got at the gas station.

A security guard welcomed me and asked me to sit down as he went to fetch the night manager. It was “old style check in” as I had to enter details into a fat log book. The room key was “electronic” (the one you swipe), though, and was handed to me along with a bottle of water and a small cardboard box. Later I found that the cardboard box had a Medimix soap, a satchet of shampoo (hadn’t seen one of those in decades now!) and a toothpaste and toothbrush.

I set an alarm for 7:30 and crashed off. IITMAA had asked me to attend the flag hoisting ceremony “around 8am”, but hadn’t given me more details. I decided to go “in search of it”, and take a walk around campus while I was at it.

After failing to find the flag hoisting ceremony, I expectantly walked towards Gurunath to find that it was no more. And having gone all the way, I went to my hostel.

Again the security guard was rather nice, and just said “oh, visitor?” and waved me past. It was 8:30 am, and I walked through the hostel for about 10 minutes “completely unmolested”. I didn’t cross paths with a single student, or even see one within 20m of where I was. The room I used to live in was bolted from inside (indicating my room-descendent was there). However, when I knocked, there was no response.

This is what my room looks like now:

Where I lived between 2002 and 2004. The graffiti, I think, is a recent addition

Most of the daytime in the hostel was spent at the end of my wing, sitting on the ledge (I’d not yet developed my fear of heights then) and reading newspapers. This is that ledge area.

Again some graffiti

I walked all round the hostel. The mess has been demolished and rooms built in its place. There is a third floor now. Large parts of the notice board in front are behind a locked glass. Even the unlocked part of the notice board has NO GRAFFITI – I guess that’s reserved for the walls now. And I was sad that I didn’t meet anyone – I would’ve loved to talk to the current inhabitants and find out what life is like there.

I had planned to meet Mohan, the legendary quizzer who was one year ahead of me at IITM CS, for breakfast that morning. Not wanting to put too much fight, I asked him to come to the guest house itself, and we ate there. The pongal and vada were good, although the “independence day special kesari bath” (a weird pink colour) was gross. We didn’t even go near the orange white and green idlis.

And then we went off on a rather long walk across campus, mainly covering the academic sections. We saw the new computer science building, and debated on what was in its place twenty years ago. We found this rather interesting nook in that building  – at the end of a corridor, a simple table and stools, and a blackboard.

The nook in the new Computer Science Block

We then went to the Building Sciences Block, which used to host the Computer Science department back in the day. And then kept walking, exploring campus and talking about lots of fun things.

It was interesting going around the place with Mohan, since we were a year apart in college and hence didn’t have any particular shared experiences, though we had SIMILAR experiences because we did the same program. This meant there was a connection but not too much nostalgia, meaning we could explore lots of different things as we walked. Oh, he recommended this book to me.

After that I headed back to my room for a quick shower, and checked out. Once again, it was time to deal with Uber / Ola. That I was deep inside IIT meant that any cab that had to pick me up had to make the trek all the way inside, and the place where I was meeting Kodhi and Aadisht for lunch was not far away at all – implying a huge transaction cost.

At least 5-6 cabs (from both platforms) cancelled on me. The ones who would call would talk about some “distance” (what I could make out given my broken Tamil) and cancel. Finally, I got an Uber that was 14 minutes away (when I booked), and which actually arrived. Turns out it came with a passenger headed TO taramani house, and I got in as he got off.

I might have written here before – I quite like going back to IIT, even though it is in Chennai! I sort of feel at home when I go back there (unlike in IIMB, where I feel like I’m invading someone else’s personal space, unless I’m there to teach or for a reunion), though I’m still very sad that Gurunath is no more.

That was the one place where I had my best memories on campus, and kept me going through my last three years there. Most of the “network” I have from IITM consists of people I hung out with there. And it pains me that it doesn’t exist any more. I really wonder what misfit students do there nowadays!

Read Part One of my trip here.

Madras Mail

Earlier this week, I was in Chennai for a day. This is the first part of my documentation of my chronicles 

A month back, Kodhi suddenly messaged asking if I can travel to Chennai for the “Landmark” Quiz, now sponsored by Zifo. “Remember 2009”, he helpfully added. That year, the three of us (Kodhi, Madness and I) had won the Chennai Landmark, and then came close to winning the “national final”.

I reversed my decade-long policy of not travelling for quizzes. I reversed my decade-long policy of not going to Chennai without a very good reason. I reversed my more-than-decade-long policy of not taking an overnight train to Chennai.

It is not like I didn’t have my share of jitters. There is something about going to Chennai that depresses me. I don’t know what it is – and that possibly explains why I hadn’t been there since 2012. On Monday night, I had finished work, had dinner at my usual time and was watching TV when the jitters came in.

I suddenly didn’t want to go. I wanted to feign illness and let my tickets lapse. I tweeted this

My wife pushed me out, saying I was being irrational and I should just go. And so I went.

The Train

This was my first Indian Railways journey since 2012. I took the metro to the railway station. The walk from the Majestic metro station to the railway station wasn’t particularly pleasant.

It was the first time in life I was travelling “2 tier AC”. I realised I’d forgotten how to climb into the upper berth. With some effort, I managed. There was a pillow and a thick woollen blanket on my berth. Presently, a steward brought a large paper bag with two (warm) white sheets and a small towel.

Then I had to pee. In the bathroom, I found a mug chained to the tap. There was also a health faucet – the first time I’m seeing one such in an Indian train. There was also a notice that we shouldn’t throw things (such as toilet paper) into the toilet. Maybe the waste doesn’t go down to the tracks any more?

I discovered that there was a curtain I could pull, to shield myself (and three others in my booth) from the corridor. Surprisingly for a train journey from Bangalore to Madras, I slept well. I started feeling less bad about going to Chennai.

At 4 am, I got woken up. Someone in the next booth had started playing Suprabhatam loudly on his phone. Then I heard someone tell him “stop it, others are sleeping”. The sound stopped. I don’t know if the two people were travelling together. Anyway, at 415, the train rolled into Chennai Central.

Indian train stations have always had homeless people sleeping in them. A new one was at Central – where I saw mosquito nets strung across dustbins, and people sleeping in them. This was just the first such example, and not a one-off.

Chennai Central at 4:15 am

The Auto

I got out of the station and booked an Uber. There was no movement in the taxi. Presently the driver called. Speaking in a mix of my broken Tamil and English, I understood he was asking me if I’ll pay cash. I said no, and he asked me to cancel.

I started booking an Ola. The inevitable thing that happens in Chennai Central happened. I got accosted by an auto driver. He initially quoted ?450 to take me “inside IIT”. I started with ?200, reasoning that it was twice of what I used to pay when I was a student. I’m not good at bargaining and I wanted to go continue my sleep, so I agreed for ?400, which was about what Ola promised to get me a cab for.

In the auto ride, I figured it’s 19 years since I graduated, and so a 4X increase in price is not that bad. I had also told the auto guy that I’ll pay him by “scan”, and he had agreed.

He presently stopped for gas, saying the lines would get longer soon. My bargaining power was low, and I agreed. Once he tanked up, he asked me to scan, for ?400 – the entirety of what I owed him for the ride. “No no, you’ll get the change in cash here. You can give that to me later”, he said. I acquiesced, collecting the ?160 in change.

At 4:30 am, Mount Road looked just like it did 20 years ago. Apart from one long and incredibly narrow flyover in Kotturpuram, nothing seemed to have changed in nearly 20 years, including the route to IIT. Oh – I saw some metro stations along the way on Mount Road.

To be continued…

 

 

4-2-4 and Huns

Last night, in the game at Stamford Bridge, Liverpool started with a formation that could have been described as a 4-2-4. While Cody Gakpo ultimately played in midfield, to make it a more conventional 4-3-3, he is ultimately a forward who was playing there, and made Liverpool vulnerable down the left side for the duration of the first half.

This wasn’t the first time Liverpool lined up in a 4-2-4 without an obvious holding midfielder. For a while during the title chase of 2013-14, Liverpool lined up broadly similarly, with Gerrard and Henderson in central midfield, and Sterling, Sturridge, Suarez and Coutinho forming a front four.

And the thing that characterised a lot of games in that title chase was Liverpool’s fast starts. I remember this game against Arsenal (I wasn’t watching) when Liverpool went 4-0 or something up very very quickly. That was emblematic of that half season – very very quick starts, lots of goals up front, and then quickly tiring and  having to hold on for dear life in the end of the game.

When Liverpool failed to score early, like they did in the game against Chelsea (when Gerrard famously slipped, and when Salah started for Chelsea), they would get immensely frustrated and look short of ideas. It was very different to recent years when Liverpool have been able to conjure up last minute equalisers and winnres.

Anyway, yesterday seemed like 2013-14 again. Liverpool was clearly the better team in the first half hour, only a very tight offside prevented the game from going 2-0. The profusion of forwards, and Alexis Mac Allister pinging balls to all parts of the frontline, meant that Liverpool dominated.

Then the inevitable happened – Chelsea settled. Their midfield three got working and soon Liverpool were massively overrun in midfield. Chelsea quickly got one back, almost got one more, and dominated most of the rest of the game (until Liverpool took of Salah and Diaz for a pair of kids).

The thing with the 4-2-4 is that it is an unusual and incredibly attacking formation. The opposition will inevitably take time to settle down against it and figure out how to deal with it. And in that time, the attacking team needs to make merry and score as much as they can (Liverpool only got one).

Once the opposition settles down, the shortage of personnel in midfield can be quickly exploited and the opposition starts dominating the game.

As I was watching, I was reminded of the Age of Empires (2; the conquerors expansion) which I used to play back in college. There, you can select the civilisation you want to play as (sometimes it’s “random”). A few people used to prefer to play as Huns.

The thing with Huns is that they don’t need to build houses (they are nomadic), and so can grow very quickly very fast. And in an AoE game, if you are playing as the Huns, the only strategy is to attack quickly and cause enough damage to the opposition in the opening stages of the game that they can’t recover after that. Because once the opposition has settled down, the Huns’ speed advantage has lost its bite.

And so, playing a 4-2-4 in football is similar to playing as the Huns in AoE. You better make a good start and inflict enough damage on the opposition in the early stages so that they aren’t able to sufficiently damage you back after they’ve inevitably settled down.

Connecting these two topics – I heard on commentary last night that Liverpool has never won a game where a Hungarian has represented them. That trend continues after last night. Hopefully Dominik Szoboszlai can make amends soon.

Creative Cycles

When you’re doing creative work, your work broadly falls into two phases – the “invention phase” and the “implementation phase”. Both imply what they mean.

There are times when you are tinkering around and experimenting to find something fundamentally new that is cool. And then, once you have made the breakthrough in finding something cool, you need to make it useful. And this can take considerable amount of work, and its own creativity.

So if you are one person doing a “creative job”, your work will alternate in these cycles – where you create and you implement. The cycles are unlikely to be periodic. Some creative solutions are so creative that implementation is a breeze. In most cases, the inspiration is only 1% of the problem – the devil in the details for which you need to perspire.

When you are part of a creative team, this cycle thing can play out in different ways. Some teams form a caste system, where one set of people work purely on the invention phase, while the other works on the implementation phase. This is especially useful when solving highly complex problems, in which case the skills required for the invention and implementation phases are different.

The big cost of having separate teams like his is the cost of communication (AGES back, when GPUs were just becoming a thing, I was part of a committee that was exploring the use of GPUs in our work. One of the findings there was that GPUs can do the work incredibly fast, but the data transfer from GPUs to CPUs was slow, and could be a bottleneck. I assume that problem is solved now). People sometimes grossly underestimate the effort involved in communicating your solution to someone else. Even if you manage to communicate, there can be significant handholding that might be required to get the other team to take forward your invention.

And so this investment in communication cost is worth it if and only if the work is complex enough. Think of large industrial projects – such as the manufacture of the iPhone, for example – they are complex enough that you need several specialist teams to perform the entire creative process. And in the larger scheme of the complexity, the cost of communication across teams is small.

On the other hand, this usage of multiple teams to perform a creative process can be massive overkill for simpler work – there the cost of communication can far overpower the gains in efficiency through specialisation.

Anyway, I’m getting distracted here.

Coming back, the alternative is to have the same people or sub-teams perform the invention and implementation stages of the creative process. Here, I’ve seen things play out in multiple ways.

Some teams are uncorrelated – this means that different members or sub-teams are in different phases of the work. As a consequence, this kind of a team constantly provides creative output. When some of the people are deep in implementation, others are inventing. And the other way round. This means that the team is constantly both coming up with new ideas and delivering stuff.

Other teams can be more correlated – either everyone is working on the same thing, or the whole team moves in sync (invention at some points in time, implementation at others). Here the issue is that there can go long periods of time without the team really producing anything – in the common invention phase, no shit is getting done. In the common implementation phase, there are no new ideas.

This can lead to stagnation in the team, and frustration outside. And so not ideal.

The other related concept is in terms of management. Some managers of creative teams are better off at managing the invention phase. Others are better off at managing the implementation phase. Given that the creative process involves both, for the team to be effective, we need managers who can manage both as well.

And this is easier said than done in a single person, and so you need a management team. And what you find is that you have a “complementary number two” (no pun intended). If you as the team leader is better off at invention, you get a number two who is better at implementation. And the two (or more) of you together manage the process.

I’ve spoken about this before – this can sometimes lead to suboptimal succession. Let’s say the inventive head leaves. The organisation promotes the implementation number two. Now, it is contingent upon this new number one to get a (inventive) number two asap. If that doesn’t happen, invention can cease. The team will carry on for a while implementing the already invented stuff, and then grind to a halt.

Similarly if an implementation head leaves, the inventive number two gets promoted. And unless a new implementation number two is hired, you’ll see lots of proofs of concept and little actual implementation. Again suboptimal.