Draw strategy

[dave]

This was first suggested last season (I can't remember who by). Whoever it was said they thought that teams who hadn't drawn for a long time were "due" a draw, and there was quite a high success rate amongst teams who had gone 8 or more without drawing. Matches where a team were in the top or bottom three were omitted (or top or bottom 2 in small leagues).
I'm going to follow the progress of it on here for a while.

[dave]

Goteberg (9 without draw) vs. Malmo (10) @3.4

(All selections 1pt)

As someone pointed out, it is possible for teams to go a long time without a draw - you need to be getting nearly 1 from 3 to break even with this plan.

I think it was Alvares who went a season without a draw - could lose you a lot of money

I really think a better strategy would be to pick draws based on your knowledge of the two teams involved.

[dave]

mick - I'm not putting money on this right now, just seeing how it goes.
Last season I seem to remember that it made a pretty good profit. I'll try and find the post for you. One team may lose you money, but hopefully the others can make up for it.

If you toss eight heads in a row does it make it more likely that the next toss will be tails? No, it is still a 50% chance. Same applies to draws, why would the chances of a draw increase just because a team has not drawn for a run of games?

[dave]

I don't know how to post links, but look at the thread "how do you pick draws?"
From 20 bets there was a 60% yield. Obviously the sample was small - that's why I'm not putting money on yet - just following it.

are what OA is talking about here Dave.

The lack of draws in the past has no influence on the lack (or bounty) of draws in the future - this is the point I was hinting at when advising you to focus on the teams rather than the history.

Statistically speaking 27% of EPL matches are draws, so picking as you are, you would expect to get 27% right <!--EZCODE BOLD START--> <!--EZCODE ITALIC START--> in the long run<!--EZCODE ITALIC END--><!--EZCODE BOLD END-->. Given that the average odds are about 3.4, you can expect to lose about 11% of your stake each time you bet, i.e. a negative yield of 11%

The fact that a 60% yield was obtained in a sample of 20 bets is a short term fluctuation about this 11%. If another 20 bets were chosen (using only this criteria) then you might just as well lose 70%!

My estimate of the 'robustness' figure here is 0.74 which is a long way away from the 0.99 that I would want to have confidence in this system.

The robustness figure (have a look in older threads) will give you an idea of how much confidence you can have that a trend spotted in a 'small' sample will be generally applicable.

i agree with what as been said.

is it not possible to flip this on its head and ask the question "why has this team drawn so many games?"

do they have problems scoring?
have they got a tight defense?
have they had a key player(e/.g striker) out?

yadda yadda...

can you really assume that the high occurance of draws for a team is independant of the perticular team?

[dave]

I'm not arguing with the fact that tossing a coin is an independent event. I'm just seeing if this works - football is a "funny old game". Maybe a team that has gone a lot of games without drawing may be losing a lot, and therefore trying harder to make sure they get at least a point.
How many games should be sampled by the way? (Obviously 20 is too small).

I'm not saying that a team who go 8 games without drawing are guaranteed a draw,

[dave]

Goteberg 0 Malmo 4

Bets W/L = 0/1
Outlay = 1pt
Return = 0pts
Profit/Loss = -1pt

I reckon you need a couple of hundred at least, did you think it would be as many?

[dave]

I was thinking about 100.

This is similar to what 'Musti' does. He's a bit of a legend for using this method of prediction, with a staking plan resembling martingale.

Can anyone explain the logic of this to me? Alot of people think this method of selection is brilliant, but I don't get it.

Dave,

Lets say the overround is 11% so if we bet draws on all games (the population) we would expect to lose 11% of our total stakes. You are being systematic in selecting your sample and you expect the mean profit / loss to differ significantly from the population, in other words you are expecting +5% to +10% edge. If you are ahead or even losing less than 11% after 100 bets that would be encouraging. I don't think that this would be a sufficiently large sample to be 99% confident that you have an edge (assuming profit). When the time comes there are some statistical tests that can be applied to derive the confidence level.

How many bets per week on average do you think you can find?

[dave]

There will probably be about 5 per week until English season starts.

Dave

if you want to achieve results that are significant at the 99% level of confidence (which is what most people would recommend as a decent benchmark), the following are the number of bets required to be confident that the given edge is a true edge and not the result of chance:

Edge Number of Bets
2% 649
5% 432
7% 346
10% 240
12% 210
15% 150

What this is saying is that if you have made 150 bets and returned 15% profit, then there is only a 1% chance that this is due to luck and 99% chance that you have selected well in some sense.

[Joe]

More excellent analysis, Sir.

Mick, how about copying that sample info into your staking plan summary in the sticky thread?

[Joe]

One other thing. What were the odds used for this analysis? If you do the stats you will find that a 15% edge with odds of 1.5 is a lot more significant than the same edge with odds of 10, simply because at lower odds you have more winners. Again it's all down to the standard deviation.

If anyone finds this hard to grasp, think about an extreme example. Bets 1000, odds 1100, 1 winner, 999 losers = profit of 100 or 10%. The statistical significance of having 1 winner in 1000 (instead of perhaps none, 2, 3 or 4) is almost nothing.

The calculations were done assuming average odds of 3.4 for draw betting.


What I might do is repeat it for home win bets and away win bets and html-ize it in the sticky thread as suggested.

M