Friday, 13 September 2013

Ratios And Risk

I posted the question of loss size : win size ratios on the Betfair Forum, and have had a few more replies.

One Forum member Darlo Bantam felt that to compare the numbers "properly" (whatever that means), the strike rate was needed, but for me the question was not how much people make, just whether my theory that an ace trader would have a close to 1:1 ratio (loss size : win size) and a less expert trader have a higher loss size to win size ratio.

I suspected that better traders would be close to 1.0 on the ratio with their strike rate being somewhere north of 50%.

Mardock posted that after 3,000 markets, his ratio is 0.82, with a strike rate of 72%.

Bayes, a much respected Forumite, posted that:

In tennis, my main trading sport, I win in about 60-65% of markets and my average win to average loss ratio is very close to 1. Four or five years ago this ratio would have been nearer 0.8 but I am much better nowadays at staying 'at market'.
His numbers for a couple of bots were reported at:
Markets 1513, Win rate 50.9%, Win/Loss Ratio 0.815
Markets 2724, Win rate 57.8%, Win/Loss Ratio 1.115
Then for tennis, and manual trading, he came up with:
All-time: Markets: 5374 Win rate 62.4% Win/Loss ratio 0.998
Pre-2010: Markets 2570 Win rate 62.6% Win/Loss ratio 0.961
TheInvestor2, another well respected Forumite, reported from the world of football:
For the first quarter of 2013, my average winning match gives me about 88% of the average losing match, and I won on 75% of matches (counting manual trading only). That was a pretty good quarter though.
Pretty close to Bayes, (and myself), and then jptrader chipped in with:
For a large number of mainly in-play football markets my average win size is about 88% of the average loss size, remarkably close to both OP and Investor.
So most traders do appear to have larger losses than wins, (wins typically ~80% of losses) and the better traders do tend to be closer to a ratio of 1:1.

There was one comment on Mark's numbers, which was posted anonymously, and a little sarcastically it has to be said:
Is that the "Professional Sports Analyst" Mark Iverson? So over 7 years he's made £200K, just have to hope his football app is bringing in the big readies cos his trading isn't.
Marks's contribution to this debate was much appreciated, and it should be made clear that Mark's numbers refer to just two sports, cricket and the NFL, and to trading activities on those two sports only.

And speaking of Mark, is he finally rethinking his strategy of easing up at the end of a calendar month? It appears so. My Casssini v Iverson post of just over a year ago (July 2012) is the ninth most popular all-time. Maybe not 'popular' but the ninth most read. What took Mark so long?

Mark is introducing a 'Market Risk Level' indicator, although what determines high, medium or low risk isn't explained. It could be volatility - certainly football would fall into the 'high volatility' category with a goal having a big impact on the outcome, but other sports like cricket or basketball have lower volatility because, (except near the end of course), any one score has less of an impact on the result. Mark wrote:
...if I’m trading a low risk event my mindset is set to be aggressive, whilst if it’s high risk I’m keen to be more cautious.
For me, distinguishing between a low, medium or high risk event is tough - at least before the event starts! If the definition of 'high risk'is an event that I am likely to lose at, then I skip it, or play with small stakes.

Play-offs certainly seem to traditionally be a challenge for me, but generally speaking I have my good sports, and my bad not so good sports, and I have no idea which games within a season are going to be good for me, and which are going to be bad for me. If I knew that...

While Mark's screenshot is the teeniest of sample, his three biggest wins come in the two low risk and one medium risk game, with the high risk games producing the smaller wins (and one small loss).  

Peter Nordsted did stop by to clarify the John Walsh selections question from Paul, explaining:
John Walsh charges me a fee for his NFL picks and I then provide these picks to readers of my newsletter which is a £15 per month subscription.
He also provides picks during the NHL season
Nick from GolfBetting sent me a message, and while golf is no longer one of my top sports, (far too much time relative to returns for me personally) he has a site full of links that some of you golf people may find useful:
Hi Cassini, Hope you don't mind but I couldn't find an email on your blog so thought I'd try this one. I thought on putting a comment on a post but thought it might be a bit spammy. I've just started a website on Golf form and have put a link on there to your blog. I know Golf isn't one of your main sports (although I seem to remember the occasional post on it). But I thought I should let you know the links there. If you have a spare minute to have a look at the site and have any thoughts, that would be great -
One thing puzzled me about the site though, was the description of this blog:
Green All Over – possibly the best betting related blog around. Mainly taken from the angle of Betting Exchanges and focusing on Football and US Team sports. Well written (and sometimes combative in style!), there is plenty that can be gleamed from the blog if you want to bet successfully.
Possibly? Combative? And I think Nick means 'gleaned' although as gleam is defined as 'A brief beam or flash of light' perhaps gleamed is just as appropriate in this case. This blog can certainly be most illuminating.

Green All Over - the betting blog that shines.


Matthew Trenhaile said...

The subject of market risk is particularly fascinating to me. If someone were to ask me during my financial services career what the risk profile of a given market was I would be over to the Options boys and their Black-Scholes dark arts to give me a gauge of the implied volatility. If I had to work it out myself during that period I would have adopted a simpler approach and studied the price standard deviation over a historic period for a given charting time period and compared it relative to the high/low ranges and the average range (possibly the median range to rule out some outliers). I would also examine the liquidity of the market because the greater the liquidity the less likely the price is to gap in market hours (what happens between market close and open is another matter). The ultimate instrument for the trainee financial trader thus being the S&P 500 futures market which barely closes and is liquid beyond imagining with strong trend tendencies and impossible to manipulate (Is quantitative easing the grossest form of price manipulation?). Back to the sports betting world and my time working for a sports bookmaker and we used to allocate the various markets we priced up a tier (1,2 or 3). These tiers would dictate the size we were happy to lay. We based our tier decision for events on the perceived strength of the market so heavily traded markets priced by a lot of bookmakers would be tier 1 and as these factors degraded so would the tier assigned it. Lots of looking on Betfair and the Asian markets to ascertain a given leagues popularity etc. So arguably everything fell upon liquidity as a deciding factor. UK bookmakers being under pressure to offer an enormous range of betting opportunities meant an awful lot of tier 3 markets and events not to mention requests for tiers 4, 5, 6 ... to be added to the scale. If you believe imitation is the sincerest form of flattery then budding new bookmakers should just look through all Pinnacle Sports’ markets and see their max bet just prior to the off to get a gauge for the global liquidity for a given betting market. The reason it is so accurate is that Pinnacle Sports need as much arbitrage volume as possible so that they as the smartest punter ... sorry bookmaker can beat their retail bookmaker compatriots into the ground. Their medium for betting ... sorry making their book is of course the indiscriminate arbitrage trader. Thus their limits calculated over time represent the likelihood of the bookmaking fraternity as a whole being able to soak up a degree of arbitrage trading which in turn is a measure of liquidity sharp/arbitrage or otherwise. So Mr Iverson could just look to Pinnacle to determine betting market liquidity or indeed record historical in play volumes from Betfair for those sports they are not so keen on. Is liquidity the be all and end all for risk assessment in the sports betting markets? Well in Betfair trading terms I certainly believe it is within a given sport and within a given format. If we were to take a variety of 20/20 cricket games then I would definitely make those with less liquidity higher risk. After all if we want to exit a trade we want to pay up as little as possible for the privilege particularly when exiting a losing trade and the less liquidity the harder it is. The harder it is to take a loss at the level of our choosing surely the higher the risk ...

Matthew Trenhaile said...

... So what about comparing risk between sports? Most sports betting markets have a trend in their price dictated by time decay but some are have moments of greater price distortion than others and these in turn are affected by the time frame as well. Compare volatility between trading a Championship football game in the first half where any goal will have a large impact even with plenty of time left in the match against an IPL 20/20 game with 3 wickets down in the second innings chasing 145 runs with 17 overs left and you have the chasing side in for 156 runs? I would argue that comparing risk across sports and across situations which may or may not arise in the course of the event is very hard indeed. On top of that surely liquidity is relative? Getting in and out of a market in £10 is sometimes easier than £1000 but probably not much easier in say a live Premier league game. So we can dictate the size we trade and the times we chose to enter and exit the market so one man trading the match outcome of the World Cup final prior to kick off in £10 versus the man who is trading in £100,000 during the penalty shoot out will engineer vastly different risk profiles for the same event. So could Mr Iverson in fact engineer identical risk levels for all events by modifying his stake size and the timing of his market participation accordingly? Maybe risk is so hard to categorise that a level stakes mentality towards it is preferable, maybe risk is simply determined not by the market but the individuals’ attitude to risk in general?

Anonymous said...

Thanks for the mention Cassini of the site. That's a bit of a spoonerism, I'll leave it in if you like and you can be the "Mr Muscle" of betting blogs! I should get around to that proof read though!!

I'll tentatively get into the debate on win/loss ratio's. Doesn't the number of trades within a market have a bearing? My assumption would be that the ratio increases the more you trade - ie if a trade is a value bet, then the tendency is wins are bigger, loses are smaller, the more bets or trades you have (assuming the trade out is at a lesser/neutral value). So isn't it as much about trading frequency than any risk aversion?

But I could have misunderstood and if Bayes is close to 1:1 then that probably suggests I am wrong! Although I would be interested to know why 1:1 would be optimum.

Reading through the argument reminded me I should put a link for Mark Iverson's blog. Although I may need some time to come up with the right type of complement?!


Grand Slam said...

Very nice blog, Cassini. Mardock is right, you have to consider the win/loss-ratio. If you go for higher odds your liability is less and the profit is higher. I think is not a good number to compare if you don't know the strategy behind. At least you have to consider the win/loss-ratio...

Perhaps you can make a link to my blog. Is in German but you can translate it.