Andreas KourkorinisTrader Interview: Andreas Kourkorinis. Photo: Stratagem

Pod Chats: Andreas Koukorinis on Sports Trading

Chat With Traders Episode 154: Andreas Koukorinis - When a quant trader enters the world of sports betting.

Andreas Koukorinis is the founder of Stratagem Technologies, a tech startup that uses AI and machine-learning technology to trade sports as an asset class.

With a global team of 80 experts in the fields of science, sports and trading, Stratagem claim that their "rigorous financial reporting and dynamic risk management allows us to offer uncorrelated returns in the intersecting space between finance and sports."

In this episode of Chat With Traders, Andreas discusses his trading background prior to targeting sports betting, how he applies his trading principles to new markets, the importance of an informational edge and the future of AI.

The World Cup or Premier League is the deepest part of the market where you could easily put a £1-2 million position on any given game.Andreas Koukorinis

On his background...

I started in 1997/98, initially as an intern for a bank that doesn’t exist right now called SG Wurburg, on the FX options desk. That was before the Euro. And then after University I started working at Morgan Stanley on the emerging markets desk and again I was an options trader, mostly derivatives and then long dated derivatives and so on. I did that for a few years. I ended up being the guy who was sort of plugging into when there were major crises around the world. After that I got given the opportunity to move to New York when the structured credit derivatives market was picking up. Around 2004 I moved to New York, for the second time, I had lived there a little while before that. I ended up trading default swaps during the first crisis of 2005 and that was the impetus of a lot of the things that I started doing in my professional life, which is basically looking at relationships between different asset classes. In this case it would be equity vs credit, equity options vs credit derivatives. I kind of basically ended up doing that for about 13-14 years.

On Stratagem...

Stratagem is essentially designed from the ground up to be a machine-learning first trading company. With a particular focus, as least for the time being, in the sport markets. We started the company originally in the early to mid-2012 and it started off in my living room with two guys who still work at Stratagem. And the idea was, can we approach the sports betting market with the same lengths that we look at financial markets? Rather than just a value-based trader. Because as a side step, most of the people who go professional gamblers, have a particular approach, which is they develop their own model and try to find essentially what they called value. Which is really a miss-pricing between the market and their model over the long run.

That difference is called the edge and they try to maximise the amount of volume they put through the market in a given period to capture that edge. We believe there are all sorts of different things you can do around that concept that come from an origin of financial markets. So rigorous risk control, risk management, take profit, looking at time series, mean reversion and the trends within particular games. All of these things have essentially the need for large data manipulation and scalable computing power. So I started off Stratagem with the purpose of building a company that will be focussed on those things. With the underlying premise that we also want to be experts in a very, very specific niche of how do you apply machine-learning in non-stationary trading environments. And that’s really what we do and that sort of evolved three things, which is; collecting lots of different sources of data and unifying them in a way that it can give you a predicted signal. developing the technology stack to deploy that signal, and then doing research around various trading algos that you put into the market once you have the data and technology stack.

On targeting sports betting...

There are a couple of reasons. So the first one is, it’s a growing market. Sports is around us everywhere and accessibility of in-play sports viewing and therefore in-play trading. It’s a 24-hour market really because you can trade a sporting event everywhere in the world. It’s fairly fragmented therefore it’s fairly inefficient, and by that I mean the concept of an exchange in the way it exists in the financial markets doesn’t really exist. So there isn’t a central clearing counterpart, there’s not a place where the price is driven predominately by the supply and demand of everybody and you can have one touchpoint to the market. The technology stack is fairly archaic and, more importantly, there is a lack of sophisticated human capital that has gone into the market.

Having said that, there are a couple of outfits and people who have been extremely successful in the sports betting space. But I think predominately the reason is because they haven’t been that many people exploring and applying rigorous trading principles as in other markets. Another thing which is important is the asset class is really, really short duration. By that I mean, if you think about a trading day, you start at the beginning of the day and you end the day with cash, so you can level yourself up multiple times without really needing a very big balance sheet to take positions. So you can be a really big trader without needing to hold a very big cash balance. Whereas in the hedge funds space, it’s kind of the other way, you basically have a large asset pool and you have to hold a lot of it as cash and you can only trade with a small fraction of what you have. So it makes the game completely different, you’re not as reliant on holding large cash balances. It really allows you to recycle things and it’s very akin to high frequency or short-term systematic trading that you’ll see in financial markets. So those were really the primary drivers for me to decide to go into sports.

On understanding the fundamentals...

I’ve been a student of trading for a very long time, maybe mid 90’s. And I think that’s the sort of the primary driver for me much more than the underlying market. So my basic principle is that if you have a trading process that is robust and scalable and as long as you do enough homework to understand the fundamental of how your market works and what drives each individual price, it doesn’t matter what you’re trading. .

On the similarities to trading...

I would say the sports betting market has various idiosyncrasies that you learn along the way and I think it’s really important to understand price formation and price action and really what does the flow of information tell you in terms of volume of trades. I break things into fundamentals, which if you make the parallel with the equity market and sports, you look at an earnings report and in our case we have analysts that write previews of games, so we want to see who’s the starting line-up, is there a particular player missing, what past performance is attributed to the line-up coming into play, have they been having a very heavy schedule, have they travelled a lot. We have our own proprietary trading models which are basically akin to the Black-Scholes model in an options world and we would look at if the bet is expensive, is it cheap and why is it cheap or expensive given the inputs in the model.

And finally I look a lot about the technical information and positioning; do I know what other smart participants are doing? If you break down the flow that goes into a sports market, it is mostly retail which for the most part does not have any information edge. Then there is a couple of people similar to us, some of them with a longer history, who we call syndicates. Essentially they are information traders, they only trade if they have an edge on the rest of the information. So we tend to decipher the flows that they drive and see if there is any information that we can glean from that. I guess one of the bigger differences that we have verses other people is we tend to look a lot at the movements of the order book and see if the price movement given the volume of trades is significant or not and is it causing moves that are disproportionate to what we expect or not? And we have our own ways and models to basically decide that information.

On bet types...

You can think about the stuff that we do as broken down into three sort of verticals. So you can do bets on a given sport, you can do bets of your own timing, they’re what we call ‘dead ball’ trades, which are basically trades that we put before the game starts and we hold for the majority of the game. This is your classical sports betting. And then you do in-play trading, which is essentially bets that you put after the game has started and you have processed the information. And the final bet is when you hold up until the end of a season. We look for who’s going to win the league, who’s going to get relegated, who’s going to win the NBA finals and so on and so forth.

And we will build a position, let’s say of the longer ones and we will put hedging instruments along the way. We also have two types of trades, one is directional trade, we have a view on who is going win the game, how many goals are going to get scored etc. or we have relative use, which is we think for example over 2.5 goals is cheap, but under 4.5 goals is expensive so we will put a position over 2.5 and under 4.5. We will play around with the ratios, wait for something to happen and then we will close the position. Or we will do things for example where we will wait for a surprise event to happen in the game, like a surprise goal, an early goal or something like that. Then look at the dislocation that happens in the market, take positions appropriately and then we will close as those things normalise.

Those are broadly the strategies we will follow. And then we will trade in all sorts of liquid markets, starting from goals, handicaps and trickle down to corners. We trade the top 26 leagues in football, then in tennis we trade ATP, WTA, men’s and women’s, everything down to Challengers and we would mostly be trading three-set tournaments and now we’re moving to also trade five-set tournaments. In basketball we trade NBA, EURO league and we're starting to trade the smaller leagues.

On finding inefficiencies in the smaller leagues...

Some of the smaller leagues tend to be more lucrative because of the inefficiencies. The problem tends to be that the scalability of your trading is not as applicable. So really there is a risk/reward between, can we be big enough to make it worth our while or is it just a small passing market? On the other hand you don’t really want to be too big or go to the really, really big markets with a dominant effect because then you signal a lot of the information that you have that others may not have. We try to have a blend between smaller leagues and bigger leagues.

Obviously the bigger leagues are very interesting because you can put big positions on. So if you think about it in absolute terms. Let’s say the World Cup or Premier League is the deepest part of the market where you could easily put a £1-2 million position on any given game. Whereas if you trade down to Australia or the US, trading somewhere between 10-15k is kind of the limit that you can trade. In basketball the NBA and EURO leagues are probably the deepest markets, so you can do £100-200k a game. Then the sizes go down pretty quickly.

In tennis I would say the grand slams are really where a lot of the volume would trade and be fairly decent on those. The trick here is that there isn’t one place where you can go and access all the volume. You’ve got to be connected to multiple venues at the same time and have the infrastructure to pull everything together to see really the global picture of the market. So building the infrastructure is really key to being successful I would say in sports betting.

On having an information edge

I would say that people don’t really have very good information, they read the paper but they don’t pay attention to the details. Which is kind of the reason why we have also been developing products for the everyday trader because we think there is a gap in the market there that should be exploited. There are a lot of inefficiencies, the bookmakers make a lot of money out of it and really it’s a market designed to be there for entertainment purposes.

So people like to place a bet when they got out with their friends, or when they sit down to relax at the end of the day etc. so they don’t necessarily take it seriously. On the flip side the majority of volume that trades is really agnostic to the nuances that go into the price information. And the bookmakers will tend to push prices one way or the other to make sure they attract people to put those trades on.

On the predictability of sports through data

The predominant factor is that sports are repeatable events. If you think about algorithms that have been developed for computer games let’s say, what’s interesting about them is the rules of the game tend to be static and the same thing holds for sports. In football you have 11 players, 90 minutes plus overtime, red cards, yellow cards and so on and so forth. So there are no factors that are extraordinary that can drive, for the most part, the outcome of a game.

And then the behaviour of the teams again for the most part as it is observable through time can be deciphered, whereas if you look at financial markets, you have micro factors that are driving the pricing of a security and you have macro factors, and you don’t really know what’s going on in the macro part of a landscape, it's really such a large amount of uncertainty. Whereas in sports events they tend to be repeatable, so if you are able to observe that is happening in a hundred thousand games then for the most part you can understand the strategies of the teams, the strategies of the coaches and how certain players play.

Want to listen to Andreas Koukorinis's full interview with Chat With Traders? Listen here to Episode 154