Grandmasterdom of Backgammon; Gambling as a simulacrum for Politics
AI and humans working as a duo; The Speedgammon World Championship and for PAID my runner-up speech
Hello Megaorderers!
I hope everyone is enjoying Summer; and the Olympics; and has gotten in some sort of holiday from politics. All summer requirements.
I am back from a working holiday of a sort in Monaco where I participated (as I do nearly every year— or more concretely 14 out of the last 17 years) in the World Championship of Backgammon.
You may wonder what skill-based gambling has to do with geopolitics or podcasts? Interestingly, I think a lot. So much of the decision matrices in our modern world are fundamentally about evaluating risk vs. reward, when to take the plunge vs. play it safe, how to hedge against risk, how to read your opponent, and how to leverage analytical models and computer analysis to make real world decisions. These are exactly what skill-based gambling is about. Coincidentally, as i’ve been drafting this piece over the last days I noticed that Ezra Klein dropped a pod on roughly this theme with Nate Silver of 538 about poker as a simulacrum for Venture Capital investment and election prediction… it is definitely worth a listen. (You can do so here) It hits on themes I’ve written about previously about in my Putin and Poker for FP or in some other articles i’ve written about gaming and politics which you can find here.
The Ultimate Backgammon Championship
Prior to this year’s World Championship of Backgammon, I participated in the UBC (Ultimate Backgammon Championship). This is an event (which although it lacks an up-to-date searchable and coherent website) is really the premier ‘streaming content’ event of the backgammon calendar. The way it works is winning the match is not the most important thing, but winning is weighted equivalently to making the right move as judged by the neural net computer programme called ‘Extreme Gammon’. (An interesting side note is that AI is based on neural net computing and neural net computing was first invented for backgammon back at IBM labs in the 1980s with TD Gammon. So it is not wrong to say that the first baby steps towards AI started with Backgammon bots.) But Back to UBC, which uses these neural nets to judge performance: I was very lucky to start this tournament with six straight match wins against one of the strongest fields that has ever been assembled:
I then finished the event in the top 8 (winning 9 out of 12 matches and garnering 7 out of 12 PR wins and posting a performance rating below 4), making the quarter final playoff as arguably the only non-professional player to be included in this prestigious knock out ‘contender tournament’ to challenge Mochy (the greatest Backgammon player who has ever lived) for the UBC title.
If this unique format of backgammon interests you or you are struggling to get your head around what it entails: you might enjoy watching the below video and you’ll find a little bit of a surprise at 2:21:00 in:
Grandmasterdom!
Over the course of this UBC adventure, XG analysis of my play in recorded matches over the last few years worth of tournament matches has determined that the moves I have selected to play at each instance were sufficiently ‘accurate’ (i.e. similar to what the computer would pick) for my overall ‘Performance Rating’ to be good enough to be awarded the title of Grandmaster of Backgammon. I am very honoured to be among the first 90 individuals to ever be awarded this title. There will certainly be many more to come. I’ll be talking more about this on the pod when I do a future episode interviewing some of the most famous BG players in the world about their political insights.
Without getting into the exact meaning of it, the title differs from the Chess equivalent by not having anything to do with winning or losing matches but only to do with the bot analysis of one’s recorded matches being of a sufficient standard (better than 4 in PR— i.e. sufficiently similar to how the computer would play the moves) over a sufficient period of time. One way of looking at the credential is that this standard of 4 PR is just about as good as the absolutely best human players in the world prior to the invention of computer neural net programmes. In short, prior to 1970, despite the rules of the game not having fundamentally changed (the doubling cube was invented in the 1920s and match play after WWII), no human was able to play at this level of performance. The game was not sufficiently understood even by the best human professionals who wrote books about it (like Dwek or Barclay Cooke).
Then, there were some breakthrough insights by Magriel, Robertie, and others and they began studying the game the way the computer programmes were going to be able to do when computing power became sufficient — that is to say they conducted rollouts (controlled experiments pitting one move against another via thousands of trials). Hence even without artificial computing power yet being brought to bear — by the late-80s early 90s, when the computer programmes were just emerging on the scene but not yet fully changing how humans thought about backgammon — 4 PR was about the top level of human play and to play better than consistently 4 PR without any advice from computer programmes was something that maybe only Kit Woolsey, Bill Robertie, Kent Goulding, and Jerry Grandell and Wilcox Snellings were able to do prior to the invention of the neural net computer programmes (first Jellyfish, then Snowie, and now Extreme Gammon). It could be said to have been the asymptote of human performance without any artificial computing help.
Therefore, the Grandmaster title means that one’s play is consistently at this World Class/World Champion level which was truly the pinnacle of human performance at this activity (match backgammon) PRIOR to human performance becoming artificially enhanced by studying with the neural net programmes on the computer. Since that interaction btw man and machine began in earnest in the late 1990s, human performance has been getting asymptotically better over the last two decades as insights from the computer programmes are unpacked and aped by human players. Why do I say asymptotically? There does seem to be an ceiling beyond which human performance cannot go. We don’t know what it is, but it might be 2PR. I very much anticipate that new young geniuses will emerge able to play even better than today’s crop of players, but I don’t imagine that humans will ever be able to play near to perfection as judged by the computer. Human mental capacity/brain power is simply insufficient and backgammon is simply too complex. This is the same reason why once computers became better at Chess and Go then humans, humans will never be able to regain supremacy. It is simply impossible for a human to be born who can play backgammon consistently in long matches below 1 PR as it is currently calculated, and this is exactly the same reason that no human (not even a better version of Magnus Carlsen) will be born who will be able to consistently beat the best computer programmes at Chess.
But like with Chess and Go, human performance via studying with machines and playing against them will constantly be improving… and this might be an excellent model for how AI should be used. It very much mirrors my discussion earlier this week with Marc Werner about how the UK government should use AI to inform and improve human decision making. Not to replace it. Humans learn principles and concepts from machines doing calculations and analysis that they cannot do but humans are needed to exercise their judgment for decisions to be made. I think how backgammon has improved the performance of human players when they play matches NOT LOOKING AT THE COMPUTER is a great model for how we can integrate AI into government decision making. Decisions are not outsourced to the AI. The AI runs simulations and experiments and it is ONE OF THE INPUTS that humans will draw on in making their decisions.
On a personal level, I am enormously proud of this achievement of being a GM as roughly the equivalent to my being awarded a UK Global Talent Visa… i.e. as an acknowledgement of being a leading participant and talent globally in my chosen field.Back go my tournament: Separately from this achievement, I was lucky enough to make the final of the Speedgammon World Championship. This is a tournament where one has only 10 seconds to make each move with a two minute bank of reserve time. You can watch what that feels like here. The final game starts at 20:50 and lasts till 27 mins and conveys the excitement and pace of the game…
Here is the match progression of my wining chances: I am over 96% 4 rolls before the end!
Nodar played amazingly well and is the consummate Georgian Gentleman so Didi Modloba to him…
I’ll end my musings by saying if Backgammon interests you at all and it is something you might want to try out… Then you need to check out Backgammongalaxy.com — play in the browser or download the app and play for free. It is the biggest BG website where all the top players play. Get your rating now and become the next grandmaster. What Galaxy provides is the ability to play human players, have your match analysed by the bots, and to learn and improve your game.
You can download the app now: link to app store: https://apps.apple.com/dk/app/backgammon-galaxy/id1606706936
link to google play store: https://play.google.com/store/apps/details?id=com.backgammongalaxy.app&pcampaignid=web_share&pli=1
Next time I am trying to train for a tournament I am going to finally follow through on my resolution to play on BG Galaxy.
And now for subscribers only here is video of a speech I gave thanking the organizers and sharing my thoughts about how the World Championships of Backgammon connect to the Global Enduring Disorder:
Keep reading with a 7-day free trial
Subscribe to Ordering the Disorder to keep reading this post and get 7 days of free access to the full post archives.