Serie Beeeeeee, i love you

[barrhead73]

Over the summer months i back tested a system which i used last season to predict Home wins in certain leagues where the home team was no more than 5 league positions lower than their opponents.

The back testing proved very inconclusive except for one league, the Serie B in Italy.

The results of back testing for four seasons were,
Season Wins/Selections
2003/04 12/22
2004/05 8/14
2005/06 9/16
2006/07 10/14

A better than 50% strike rate with average win odds of 2.16. The extra selections for 2003/04 can be explained as the division had 24 teams in it rather than 22.

Obviously with only an average of 15 selections per season you are not going to become a millionaire, but i will run with this system throughout this season to see if makes a profit again.

Need to wait until 8 games have been played so there won't be any selections for this weekend.

Selection criteria is based on league position, last 6 games, last 4 homes & aways and the opening odds.

Still to decide on a staking plan, but will let you know before the first bet is placed.

[aliando]

hope it continues for you mate!

[barrhead73]

The system throws up a game for this weekend,

14 Oct - Grosetto tb Messina @ 2.3 (Bet365)

Stake £3

Starting Bank is £300
Staking plan will be Rising Floor Plan

[phantomDude]

I dunno, a system that only throws up 16/17 bets a year? And which only has a success rate of slightly higher than 50%? And you've only backtested it on a couple of dozen matches?

If you're happy placing that few bets, fair enough, but I don't think it's that big a deal. You've tested the system on approx 60 bets, and not made a massive profit. If you picked batches of 60 matches at random out of a database, and backed the home team to win in all of them, you'd be surprised how many times you'd be in profit (and sometimes significant profit). This would be true even when the return from betting on all the matches in the database is negative (ie approx 10/11% loss if you took odds from high street bookies).

By all means, if this is something you want to investigate, do so, and have fun. I just don't think that the returns from your analysis are spectacular enough or that you've used a statistically significant number of matches to perform that analysis.

Anyway, I'll stop being a nerd now. Enjoy your bets!

[barrhead73]

Phantom

I think you misunderstood my first post.

I have back tested the system on every match from Serie B over the last 4 seasons. Therefore this is not 60 matches, this is somewhere in the region of 2,000 matches (a slight difference). The system threw up 66 matches and 39 of them won at average odds of 2.16.

I also back tested this system in 8 European Divisions but found that only Serie B gave me a consistent result, thats why i'm only going with this.

But if you fancy a little light hearted challenge then lets go!!!!!!!
Pick a home team with odds > 1.9 (Only on Bookies sites, no exchanges) for this weekend. I shall keep a record of how we do.

[phantomDude]

I stand by what I said.

I said that if you placed bets on 60-odd games picked at random out of a pool of thousands, you would (many times) come out with a profit. I don't think it's significant that you combed through 2000 games, picked out 60 bets, and these showed a profit for you.

If you said you'd placed 2000 bets, and you were showing a profit, that would be significant. Or, if you'd picked 60 bets at odds around 2.16, and *at least* 80% were winners, that might be significant (only might, due to the small size of your sample).

I've analysed lots of strategies using historic data, and coming out with a profit on the scale you're getting over 60 matches is nothing unusual, most times when you look at how the system performs over a larger number of matches, you'll have patches like this, in among long periods of much lower returns.

So, in short: I don't consider your system to be tested over 2000 matches, it's tested over 66. The size of the potential pool of matches you use for testing your data is irrelevant; what matters is the matches you actually place a bet on.

I don't know about the challenge, you can blast away yourself if you want, but I'm pretty busy at the moment. If I get time, I'll post up a few selections, but I can't guarantee anything.

[rich46]

Take the challenge Dude, just for the fun of it May the best man win. If you do the best of luck to both of you!!

[Matthew]

I agree with phantomdude. Although I'm nowhere near as critical.

Yes barrhead73 may be a little over zealous in his claims, but if we only
had threads which were proven over 1000s of picks then we'd onlly have 1
or 2 running and certainly no new ones. This place would soon stagnate.


Barrhead73 I see there has been 8 rounds in Serie B now, can we expect some picks?

??? Ascoli(8th) v Chievo Verona(5th)
??? Grosseto(16th) v Messina(12th)
??? Ravenna(18th) v Triestina(14th)

[phantomDude]

Critical??? Me?

Better watch what I say around here!

[barrhead73]

Fair enough Phantom, i take your point about the level of bets tested, but i feel my testing has been done on a lot more than 66 matches.

However, your point about just choosing random home wins and winning big profits? I find this very strange that every punter here is not in profit then.

As i said before, please feel free to pick random home wins and post them up on here and we shall see how our systems go.

I did state that even if this system returns the same as previous seasons we won't become millionaires but i like to think that by using this system and also a few others on here, then hopefully a good profit can be made each season.

I'm currently following JTW on his Unders system, Merlins Michael Wray picks, Fleets Trebles and Matthews Premier Picks.

Matthew, nice to hear from you, my pick from your 3 games is highlighted in message number 3. The Ascoli game just narrowly missed out.

[phantomDude]

I have a database of over 70,000 matches. If I were to bet on the home win on everyone of those matches, I'd be down approx. 9% on what I'd staked (ie the bookies overround).

I'm selecting the matches from that database, in no particular order, with a column indicating what my winnings would be if I just blindly bet on the hometeam, and I'm picking out winning streaks of 70/80 bets where I'm up 10/15 units for fun. They're everywhere.

Here's an example to illustrate my point. I divided up the bets into sets of 300, and I kept track of what my profit/loss was for each set of 300. Here are the results (over 3000 bets):

(Batch,Start bank,End Bank,Profit)
1 0 -49.38 -49.38
2 -49.38 -49.13 .25
3 -49.13 -83.04 -33.91
4 -83.04 -110.17 -27.13
5 -110.17 -134.19 -24.02
6 -134.19 -197.89 -63.7
7 -197.89 -214.71 -16.82
8 -214.71 -240.1 -25.39
9 -240.1 -235.21 4.89
10 -235.21 -219.34 15.87

As you can see, 7 out of 10 of the sets of 300 gave negative returns. One gave virtually no returns (.25), another returned 4.89 (a return on your stakes of over 1.6 %) another returned 15.87 (a return of over 5% on your stakes). Even though over the long term (70,000 bets) you will lose 9% of what you stake, there was significant variance over smaller numbers of bets (300 bets). I'm not a maths genius, this is something to do with the Binomial Distribution, but basically my point is that performance over 60 bets, even 300 bets, is not necessarily indicative of how a strategy will perform ad infinitum. In the ten sets of bets here, we've seen returns varying from gains of over 5% (which, from a bookie, if you could maintain that persistently would make you a millionaire very quickly) to losses of over 21%. And the only thing that differed between those sets of bets was the natural variance that occurs in any stochastic process.

My point about placing bets at random was that it's perfectly possible to place bets in a fashion which makes you money over the short term, but will ultimately go on to lose you money if you persist. I'm not talking down to you, I remember being baffled as to how a system which on paper had a 60% strike rate resulted in 1 win out of 8 for me. It was only when I crunched the numbers a bit more when I learned that the returns from betting follow what resembles (to borrow from finance) "a random walk", albeit a walk that is inclined downwards because of the bookies overround. They most certainly do not follow a straight line!

Anyway, lecture over. I'll place a bet or two here when I get a chance.

And there's no point in me picking out random home wins, my returns will tend towards -9%(ish). I'd bet me right bollock on it, it's a near mathematical certainty, but *only* over a stupidly large number of bets.

[barrhead73]

Excellent explanation Phantom.

I can only hope that my system will prove you wrong.

[Matthew]

Originally Posted by barrhead73

Excellent explanation Phantom.

I second that.

Barrhead73: I didn't realise you followed my picks......apologies for the god
awful start to the season. I hope you persevered and have made a few bob in the last couple of weeks.

[barrhead73]

Yes i have followed all season Matthew and followed for the last 3-4 months of last season.

Excellent thread, and i fully expect your fortunes (and mine) to turn around soon.

Hope i can have some success here with this as well.

[barrhead73]

Grosetto 0 Messina 0



Bank now £297

[barrhead73]

Nothing for this weekend.

[F.]

phantom, your explanation post was simply fantastic. Keep up the quality posts.


F.

[barrhead73]

Nothing for this weekend

[barrhead73]

Nothing meets the criteria for tomorrow nights games.

[barrhead73]

Nothing for weekend of 3rd November.