[Brian72]

Hi, apologies if this is covered in another post. Please direct me there if it's the case. I did a search, but found nothing.

OK, I have a system that crunches tennis results from past years and churns out a prediction between a matchup based on those results. I get over 70% correct results with this system. Slightly better on ATP than WTA and better on Futures ($10,000) and ITF $10-15K womens events than on higher level tournaments. It ignores form, which is something I need to rectify. The exact percentage varies from week to week, for example, it dipped a bit bit when the European clay court season began.

Unfortunately, it's hard to get good odds or any odds on futures and itf tournaments. I have tried to get as much odds data as I can from the net to back test predictions, and it says I'll make a nice amount of money based on the odds (which are incomplete and so maybe those games I don't have odds for are not randomly distributed and skewing the results) but when I bet the amount suggested using half-Kelly instead of small fixed amounts, I don't have much success for any length of time, lose most of my bank and have to grind a bit on fixed amounts to get my bank up again.

I guess my questions are, is +70% enough and if so, is a Kelly type betting system the right way to go?

Thanks for any help. Apologies again if this is in the wrong place or has been answered.

Oh, and what is a good starting amount for a bank?

[Hardcoar]

Who the f*ck is Kelly?

[Cessation]

Who the f*ck is Kelly?

The bookies second best friend behind Gale, Martin.

[faststeady]

its a bayesian thing hardcore, those lock monsters lurk in team sports, this person kelly is one of them

[HeeeHAWWWW]

It's probably not too difficult to beat books playing futures, but the books that allow you to bet there don't tolerate winners. You'll find yourself booted or limited to nothing in no time. That used to be the case for challengers also, but pinnacle cover those now, although at fairly low limits. You can perhaps get better odds elsewhere at times, but again, beat em and you'll lose the account.

Wrt Kelly: it'll always increase your bankroll faster than flat betting, unless you are on average over-estimating your edge. Can't over-emphasise that. Beyond an optimal stake you're reducing your bankroll growth and the variance skyrockets, and beyond 2x you're actually going to lose money longterm, even on +EV bets.

That's what theory says, and it's also what I've found over 4500 bets now (I record my +EV predictions for analysis later). Half-Kelly is probably too aggressive though, I use 1/3 and the up/downswings are still brutal. A rather large pro syndicate a while back revealed they settled on 0.4xKelly, but tbh I think that's too high for tennis.

Another element regarding Kelly staking in tennis: you might have a gigantic number of bets, but much or most of your profitability depends on a small subset of high edge bets (which of course with Kelly are the largest stakes). Your accuracy on these ones is critical, any over-staking there is going to destroy you. It might be worth experimenting with capping your estimated edge, or applying some sort of crude limiting handbrake - eg multiplying it by 0.99^(ev*100)