Tuesday 26 January 2016

Stakes Well Done

My comment was that Kelly was the optimal staking system, adding that:
To suggest that variable staking 'can' leave you exposed with just one bad bet is technically true, but it's also meaningless. It's idiotic staking that can hurt you, not variable. No single stake, whether level or variable, should ever deviate far from around 2% of the bank, perhaps up to 4% as detailed in the link - unless you have inside information of course.
Webbo responded with these thoughts: 
Of course agree that Kelly is the optimal (we use it for the Holy Grail in fact and showing a decent profit yet again for this season :) ) and what we were trying to get at here is that if you are betting at level stakes you benefit each time the bookie is wrong. You can afford to be wrong more often.

But if you are using Kelly or a similar variable method, you need to make sure you don't get your pricing drastically wrong on ANY bet.

Find one even money shot you think is a 95% certainty and you've just wiped out a massive chunk of your bank. An extreme example maybe but even if you are not 'overstaking' one badly priced bet can still undo the good work of several other good ones.

Letting the bookie do all the pricing will work better for 99% of punters. Unless you have a computerised system that prices up events for you like we do.

I should think bookies probably hate large level stake punters more than almost anyone else.
It's probably easiest to reply to this one point at a time.
Of course agree that Kelly is the optimal (we use it for the Holy Grail in fact and showing a decent profit yet again for this season :) ) and what we were trying to get at here is that if you are betting at level stakes you benefit each time the bookie is wrong. You can afford to be wrong more often.
When the bookie is wrong, and by "bookie" Webbo means the market, the punter benefits with ANY stake, however the size of the stake is calculated.

It's also worth mentioning that it is the size of your level stake relative to your bank which is critical. If your bank is £1,000, a level stake size of £500 is not a good idea. It also makes sense to compound your edge by increasing stakes as your bank builds. Level stakes aren't going to do this for you. If you have a more sensible level stake of say £20 for the above bank and are skilful enough to build your bank up to £10,000, your level stake of £20 no longer makes sense.
But if you are using Kelly or a similar variable method, you need to make sure you don't get your pricing drastically wrong on ANY bet.
The key word here is "drastically" which means "acting with force or violence; violent, extremely severe or extensive".
Find one even money shot you think is a 95% certainty and you've just wiped out a massive chunk of your bank. An extreme example maybe but even if you are not 'overstaking' one badly priced bet can still undo the good work of several other good ones.
That is indeed an extreme example, and in line with the "perhaps up to 4%" philosophy, anyone considering risking 90% of their bank on any one bet is probably not well suited to betting. Assuming no inside information, it is frankly delusional to think that you could have anything other than the smallest of edges over the market.

The beauty of Kelly (or more conservatively fractional Kelly) is that if you are able to measure your edge, you can maximise your returns, and your bank builds faster than it does with a flat level stake.  

The curse of Kelly (or more conservatively fractional Kelly) is that accurately calculating your edge on a sporting event is very difficult. For that reason, sticking to a maximum bet size of 4% of your bank will serve most people well.
Letting the bookie do all the pricing will work better for 99% of punters. Unless you have a computerised system that prices up events for you like we do.
I'm not sure I understand this statement. The market determines a price, and it's usually pretty accurate. Unless you calculate your own probabilities on outcomes, how are you ever going to beat the market?
I should think bookies probably hate large level stake punters more than almost anyone else.
The bookies have no idea whether your stake is a 'level' stake, a fixed percentage stake, a Kelly stake or whatever - and they will soon stop you betting any stake if you are a consistent winner! If bookies hate anyone, they close or restrict them.

4 comments:

James said...

I agree entirely. If the other side of the trade has negative expectation then you have the edge, regardless of stake. Edge and stake are not related in that respect.

Poor money management affects the variance in profits and level stakes is just as likely to destroy your bankroll by yielding less profit on a winning bet as too big a loss on a Kelly bet.

The optimal bet is to chase your winnings and not your losses. If you consistently bet less than you should then you are building a smaller cushion for when the inevitable losing streak comes.

Kelly betting conforms to "chase your winnings not your losses" by sizing the bet to your current win/loss streak. If you are in a losing streak then Kelly will make smaller bets relative to the decreasing bankroll.

Brian said...

Largely agree with you of course and I'll just clarify a couple of the points.

By letting the bookie do the work, I mean let them decide on the exact price. It's much easier to know that a price is too big than to know by exactly how much each time.

I'm sure bookies do hate all winners but most of those who use a variable method are less likely to be able to price up markets as accurately as the bookies on a regular basis. Of course you could be varying based on odds ranges only but this is then not much different to level stakes if that’s the case. They will also likely see more volatile swings of which they are less likely to be able to handle. I’m guessing a bit here as I’ve never been a bookie and I’m not saying that I’m correct, but after an early period of success I think they’d be more confident that the variable staker will come a cropper than the level stakes bettor.

Of course there will be some exceptions but very few people will be able to show a set of results that have been improved by variable staking.

Brian said...

We are all singing from the same hymn sheet here by the way. We all know that Kelly is the optimal approach in theory but 90% (maybe more?) of punters won't be able to use it or any variable staking to their advantage. The twitter statement was merely intended as advice for most people.

Anyway here's a more interesting question for you related to this. How low do you let your bank get using Kelly if you have an edge? How low would you expect it to go?

In your mathematically perfect world you would utilise as much of your edge and bank as possible but what if these sees you lose up to 40% of your funds along the way?

Our simulations over more than 10 year’s worth of data show that if we make the most of our edge in all markets we could see our bank more than 40% lower than the previous bank high and it can take up to a year to see a new high.

This does return by far the most profits in the long term but when the time comes for us to drop such a percentage, how do we know there's not some external forces that have come into play or that the edge has diminished? At the same time we want to maximize our edge as quickly as possible.

Appreciate your thoughts.

James said...
This comment has been removed by the author.