Read on Twitter
1/ Bankroll management tends to be an afterthought in MMA betting conversations, but it's one of the most important concepts. Even a highly profitable bettor can go broke with poor bet sizing.
Here's a thread with my thoughts on the issue:
Here's a thread with my thoughts on the issue:
2/ To start with, rid yourself entirely of the idea of a betting record (in terms of number of wins/losses) mattering at all. Same with number of profitable events.
The goal is to optimize for profit, not for a meaningless "win rate".
The goal is to optimize for profit, not for a meaningless "win rate".
3/ If you make 0.5 units of profit for 25 events in a row then lose 20 units on a single freak upset, you're still in the red. You've been doing what Nassim Taleb describes as "picking up pennies in front of a steamroller", and your "win rate" is just a useless marketing gimmick.
4/ Yes it's easy to say "well that was a freak upset who could have seen it coming", but that's not a valid excuse when we *know* freak events happen in MMA.
(Side note: read Taleb. Tons of concepts applicable to sports betting)
(Side note: read Taleb. Tons of concepts applicable to sports betting)
5/ Conversely, if you specialize in long-shot prop bets and are often betting +700 lines, win rate will look terrible. That doesn't matter, because you're taking lots of little losses and more than making up for it with big wins.
Profit is what matters, win rates are snake oil.
Profit is what matters, win rates are snake oil.
6/ So if we're focused on the long-term magnitude, not the frequency, of wins, lots of common betting patterns are terrible.
Betting 0.1u on all your long-shots and 5u on all your heavy favorites is going to give you a great "event win rate" and destroy your bankroll.
Betting 0.1u on all your long-shots and 5u on all your heavy favorites is going to give you a great "event win rate" and destroy your bankroll.
7/ Bet sizing isn't just about the risk and potential payout, it should also depend heavily on how wrong you think the line is (aka your "edge" on a given bet).
A +2000 line that should be +1000 shouldn't be getting a bet 5-10x smaller than a -1000 line that should be -1200.
A +2000 line that should be +1000 shouldn't be getting a bet 5-10x smaller than a -1000 line that should be -1200.
8/ The concept to start with is the Kelly Criterion. Graph here from Euan Sinclair's "Volatility Trading".
This displays profit/loss paths of the same 'bets' (trades) made at different sizes. More on half and twice Kelly in a bit.
This displays profit/loss paths of the same 'bets' (trades) made at different sizes. More on half and twice Kelly in a bit.
9/ The Kelly Criterion maximizes profit by factoring in the potential payout, your bankroll and the 'real' odds of an event happening.
In sports betting, these "real" odds are where you're capping a line.
In sports betting, these "real" odds are where you're capping a line.
10/ With Kelly, you're re-calibrating your "unit" size with every event, as your total bankroll (and thus the Kelly optimal bet size) changes each time.
In theory (ignoring min. bet limits here for sake of example), a profitable bettor can't go broke following Kelly sizing.
In theory (ignoring min. bet limits here for sake of example), a profitable bettor can't go broke following Kelly sizing.
11/ This means that you will survive and have enough money to capture your true edge over the market in the long run.
"Long run" is the key though, notice that in the graph above, 200-300 bets in we're not seeing a huge discrepancy in the results.
"Long run" is the key though, notice that in the graph above, 200-300 bets in we're not seeing a huge discrepancy in the results.
12/ The volatility and the swings suck. Look around 701-801. If you make 5 bets per event, that's a 20 *event* downswing that sees you lose about 2/3 of your bankroll.
If you get discouraged and don't stick to Kelly in that valley, you could make things much worse.
If you get discouraged and don't stick to Kelly in that valley, you could make things much worse.
13/ So we need to factor in the human element here, including margin for error on our capping.
If you have a 20-event, 67% of your bankroll downswing, you're not going to have any confidence that you're still a profitable bettor.
If you have a 20-event, 67% of your bankroll downswing, you're not going to have any confidence that you're still a profitable bettor.
14/ For practical purposes, this means we want something more like "Half Kelly", which is just betting half the Kelly optimal size.
Even the best cappers in the world won't really know if -250 or -255 is the right line for a fight. Using Half Kelly gives you more wiggle room
Even the best cappers in the world won't really know if -250 or -255 is the right line for a fight. Using Half Kelly gives you more wiggle room
15/ Also notice what happens with Twice Kelly (betting 2x optimal) on that graph. You go broke making the same bets that could be making you profitable.
Like I said, if you bet 10 units on a -1000 line when you should only be betting 5, you're going to go broke eventually
Like I said, if you bet 10 units on a -1000 line when you should only be betting 5, you're going to go broke eventually
16/ Bet sizing really, really matters.
There are plenty of Kelly calculators online, this is the one I use: https://www.albionresearch.com/kelly/default.php
Whether you're risking full Kelly or doing half or even quarter Kelly sizing, be consistent.
There are plenty of Kelly calculators online, this is the one I use: https://www.albionresearch.com/kelly/default.php
Whether you're risking full Kelly or doing half or even quarter Kelly sizing, be consistent.
17/ Optimizing with a slightly less profitable approach you can stomach and commit to for the long term is going to make you more money than trying to use full Kelly sizing but changing your mind once it means making a single bet for 50% of your bankroll.
