131. Moneyball 2.0 (Optimizing Labor)

Moneyball (2011) transformed business, despite it containing misinformation. 15 years later the world has changed, and it is time for an update.

I grew up very poor and was homeless from age 17 to 22 on the streets of Los Angeles. Despite this I had very ambitious goals which I met, and in order to do this I had to maximize efficiency everywhere in my life. 5 minutes lost could cause my personal supply chains to break down and mean not getting enough calories that day or getting a “B” on one of my university exams. Thus my favorite word was “efficiency”. My friends would make jokes about me being an action doll and it would say that word if you pulled a string. It is fair to say I had an efficiency fetish.

Thus when Moneyball came out, and envisioned a world where you could more efficiently hire and organise people to get jobs done, it certainly got my attention. I wrote 3 papers on Moneyball:

• Moneyball, Zynga, and other Unconventional Applications of Economics (2012)

• Moneyballification (2012)

• Six Years After Moneyball (2017)

I have written a number of addition articles focused on improving team productivity more recently, especially when I focused on interpreting the results of the massive Google Project Aristotle in my Corporate Merit paper (2024).

When I recently analysed the research conducted by Electronic Arts and NetEase on matchmaker technologies it bothered me that both companies were not using their labor assets in an efficient way. This resulted in the research costs being very high, and the productivity/quality being relatively low. Now with AI labor involved, it is time to get into a brutally objective analysis of labor allocation in supply chains as it is no doubt time to reorganise those chains. Some tips could save your company.

For the purpose of keeping this simple, I will classify labor into 3 categories:

1. Multipliers (Intelligence)

2. Adders (Production)

3. AI (mixed support)

Adders are the simplest category so I will start there.

Adders

In the old days you might have 8 farmers getting X work done per day. If you doubled that to 16 farmers you would get 2X work done that day. Each additional worker added the same work as the previous workers. Thus if you needed more work done, you added more workers (adders).

In Moneyball, a scenario was created where you needed to get the most work done possible but you had a limit of 9 active players on a team and their individual skill levels mattered. But not every position had the same skills. Hiring a player that was good at everything was expensive and it was cheaper to hire players with varied (but incomplete) skill sets and have them teach each other. In this case, 3 incomplete workers could be cheaper than a complete worker. Since the number of workers here was capped, the focus now was on the most efficient allocation of skills on a set budget.

Companies still try to hire complete workers. Almost 100% of the time. So they didn’t learn anything from Moneyball, and that came out almost 15 years ago now.

Multipliers

Also known as intelligence workers, or expert workers. Having complex knowledge available to a team can multiply the output of the entire team. Having one person double the output of the entire team is entirely possible and is normal for a properly positioned multiplier.

For some supply chains, the absence of an expert can drop production to close to zero. In this case, having that expert in place, versus not having one, could be an almost infinite multiplier. In that case, that worker becomes essential to the supply chain. If you fail to get an essential worker, or if you lose that worker, the entire production might have to be shut down permanently. This happens when the skill sets provided are extremely rare and/or in high demand.

This is what happened at Wargaming when I was laid off because I was American. Leadership was told I was an essential worker but were acting from ideology not productivity. So all the supply chains on 5 games on 3 continents essentially shut down or were given busy work. For years. A similar situation occurred, after my being given an illegal order by a 23 year old, in a company that failed to identify my status as essential prior to wandering into unprofessional territory.

Non essential multipliers can be dismissed at will but if they are a 2X multiplier your supply chain will drop to 50% efficiency until this person is replaced. Multipliers can vary in effect widely, which is not always understood by adders. Management is often production (adders) promoted to management due to seniority. They can take on a multiplier effect because of their global understanding of the supply chain and ability to organise others, but they don’t generally think like multipliers.

Having two multipliers of the same type (say 2 biologists) does not improve the multiplier effect. If you have a 2X biologist and a 2.5X biologist, you will operate at 2.5X efficiency. If you have six 2X biologists and one 2.5X biologist, you will still operate at 2.5X efficiency. Multipliers are not adders. So when adders (like HR people) hire scientists (like computer scientists) they often make the mistake of thinking “oh, 7 computer scientists is better than 1” like both the aforementioned Electronic Arts and NetEase teams.

They would have been more productive with their top CS multiplier and a 1.5X psychologist or neuroscientist. This combination is also a LOT cheaper.

Here’s the rule: Adding more adders boosts production. Adding more multipliers does NOT boost production. You just need one of each relevant type. Adding additional multipliers that are in different fields that are complimentary CAN act as an additional multiplier. In this case the effects are cumulative. You only need more than one of one type if you think your primary could die (often from old age, they are experts for a reason), or take ill, or might leave for some reason. Then it’s good to have a backup that already knows your supply chain. If you have an essential worker, then having two of those would be a very very good idea. But that’s typically almost impossible.

I’ve had an employer try to get me to train my replacement (this is pretty common all across industry) after I refused that illegal order. This is a bit rude, and very unlikely to go unnoticed. Also, transferring knowledge from an essential worker with 12 years of university work, 15 years of independent research, and 25 years of industry experience to another 20 year old, would probably take more than 10 years. You aren’t going to make that happen in a month, especially if the relationship is antagonistic.

You may have noticed that AI can now out perform ~80% of scientists, in their field of specialty. That’s an impressive number. But if you’ve read my paper on Scientific Cynicism then you might realize that science today is dominated by nepotism and politics. There is a HUGE difference between the bottom 80% and the top 20%. Just having a PhD doesn’t tell you much about the capabilities of a person. There are no standardized tests for PhD, it just means they published and defended one research paper. That could have been on black holes, or the dating habits of prairie voles.

Thus the bottom 80% might be between 1.0X and 1.5X multipliers, and the 99th percentile could be 10X or higher multipliers. Generally the top people in each field are known by name by almost everyone else in that field and have extensive published research histories going back decades. I don’t mean to belabor this. If you are in the field of science you are well aware of all this. But for laypeople this can all be a bit mysterious.

I realize that game development companies generally have zero or almost zero multipliers/scientists employed. Computer scientists are really more like engineers than actual scientists, but are still multipliers if they have relevant complex knowledge not known to regular computer science graduates at the B.S. level that improves the performance of everyone else on the team. Like all engineers.

This situation is going to need to change if larger companies want to remain competitive in an industry that is increasingly going high tech.

Advanced industries like pharmaceuticals, processed foods, energy sector, and military will have entire hierarchies of scientists, with lower levels watching specific supply chains and the top ones creating new trickle down technologies. But game development just has not gotten on the science bandwagon at all.

AI

The point of Moneyball, which I refined in my various papers, is that you want to offload the more mundane tasks in the supply chain to less expensive workers. At the same time, you want enough expert diversity to have at least one multiplier in every position that can accept a multiplier.

So for instance if a role needs advanced scientific skills and also excel or data science expertise then what you have here is a mix of requirements. The scientific skills might take 10, 20, or even 30 years to learn. The excel and data science skills can literally be learned in 6 weeks. If your supply chain needs a lot of excel or data science work then you should hire multiple dedicated adder employees in these roles to free up your more expensive multipliers to focus on the work only they can perform.

Game development is failing to do this, which is unfortunate because even that level of productivity is now obsolete. Now with AI most of your adder work can be done by AI. So now that team of 6 excel techs and data scientists could be done with one person who knows how to ask AI to do this for them. Or one person who is good at the technical skills, and another whose job is just to ask good questions of AI. Your old fart multipliers might be a bit too low in neuroplasticity to be learning these new skills on time, so this is an ideal role for young hires.

Even for multipliers, AI is really speeding things up. Research typically involves knowing what has already been done by others before trying to figure out what new research needs to be done. AI speeds up this sort of search and can help find research that has been published in odd places.

Since I’m a rogue scientist (not entirely by choice) my work often gets published in “odd places”. So I understand if these matchmaker research teams were not aware of my prior work in this field. Missing that cost their employers at least tens of millions of dollars. One of my long time followers messaged me yesterday to tell me that google is already putting the paper I published yesterday regarding EnMatch on page one of related google searches. If you searched a traditional research database, it would not show up. These MM research papers also might not have any peer review because they did not follow standard academic procedures. Thus my peer review might be all that is out there.

In the last year or so as I’ve been using AI I can now do that prior research query in minutes. This used to take me weeks. So instead of putting out a paper once a month, now I can publish novel work in one or two days.

Rebuilding Supply Chains

All this change means that not only are pre-Moneyball supply chains (80% of what’s out there) obsolete, but even post-Moneyball supply chains are (the other 20% of what’s out there). The result is that you could get 2 or 3 times as much productivity from half or less the number of employees you had before.

I would expect a lot of adder positions to disappear, along with redundant multipliers (you don’t need 7 of the same kind of scientist on one project). A more diverse array of multipliers can then be employed, and there are going to be a lot of new jobs opening up. Like AI liaisons whose job is to ask AI for work outputs.

Some teams, like the one at DeepSeek (in China) had to do all of the above because their supply chain was under attack by a hostile foreign power (the USA). They didn’t have the luxury of not rebuilding their supply chain. But in doing so they ended up with a higher degree of efficiency than their competitors.

It’s unfortunate that many companies will not rebuild their supply chains until they fail entirely. But if you look at AAA right now, how can you not see that the supply chains are collapsing across almost the entire sector? It’s so bad that even game hungry consumers are asking for game developers to be fired. They are still desperate for games, but even laypeople can see that development resources are being misallocated.

For decades Hollywood has had predetermined prebuilt supply chains for almost any project. They know where to go and who to call. They can raise the team they need, do their shoots, and then dissolve the team rapidly. Because hiring and firing on short notice could harm workers, they have well established unions to reduce the risks.

AAA is decades behind Hollywood in this regard. Given that AAA is more tech dependent, and that tech is changing so rapidly, they really need to be ahead of Hollywood by now. I’m not sure what they are waiting for. As Hollywood attempts to replace workers with AI, this is disrupting the careful balance of power and unions are reacting (appropriately) with some alarm. Eventually the old formulas for their supply chains will be completely rewritten. That’s what all the lawyers are arguing over right now.

Hollywood can get away with rebuilding their supply chains perhaps once every 20 years. Right now in game development that seems like it has to be done at least once every 5 years. Some projects last more than 5 years. So those projects end up being obsolete before they can even get to market. Better supply chains should be able to deliver games in 2 years or less. At Wargaming, even with relatively green teams, our time to market was under a year on all products. But then because those were new teams, they were also new supply chains. Thus they were able to out perform much more veteran studios that had exceptional talent but were mired in inefficiencies.