Forums

General Betting

There is currently 1 person viewing this thread.
Cuban
29 Feb 12 15:30
Joined:
Date Joined: 11 Jul 01
| Topic/replies: 33 | Blogger: Cuban's blog
All, I am currently testing a system I have created using small stakes.

The system places outright back bets. So far I have placed 215 bets, average odds of 3.4. My net profit (after comm) is 35 points, yielding 16% ROI. Strike rate is 38%.

At what point does the test become robust enough to start increasing stakes? I'm sure there are some clever mathematicians on here who can offer advice.

Post your reply

Text Format: Table: Smilies:
Forum does not support HTML
Insert Photo
Cancel
Page 1 of 2  •  Previous 1 | 2 | Next
sort by:
Show
per page
Replies: 74
By:
steeringjobnap
When: 29 Feb 12 18:33
for n>= 30, the sample's average will approach the mean of the population being sampled as n increases towards infinity.

Personally speaking, I will define my expected value-per-bet before each wager, so am not too pushed about such sampling -> E[x] will always > specified margin before playing.
By:
loserschaselosers
When: 12 Mar 12 19:55
in excess of 1000.
By:
kenilworth
When: 13 Mar 12 09:59
I think it is most important if you are
including an logical ingredient seemingly
overlooked by others, but 200 picks, strike
rate, ROI as stated, seems good enough.
2 Other points, I personally don't like paper
trading or back fitting, beware of those.
By:
DivideByZeroError
When: 13 Mar 12 22:06
Look at the P&L for each race and calculate the mean and standard deviation to do a z-test: http://en.wikipedia.org/wiki/Z-test

This can give you some confidence that the profit isn't just due to chance.
By:
getintheir
When: 13 Mar 12 22:49
these stats guys will always say you never have enough data.
By:
TheInvestor2
When: 13 Mar 12 23:19
Are you offering or only taking bets? If you're offering, increasing stakes will likely reduce ROI due to getting more matched when are wrong than when you are right.
By:
kenilworth
When: 14 Mar 12 09:53
I think 200 results is good enough if not back fitting.
By:
DFCIRONMAN
When: 14 Mar 12 11:12
Just "curious" as to the sport you are betting on?

If HORSE RACING, then you can get runs of "good luck" and runs of "bad luck" that can distort results. So the more you check results, with thought about them, then the more it will be clear that a real "edge" does exist.

200 seems far too low to be "confident" re future events....if it is HORSE RACING. I've checked almost 2,000 races, so have a "deep" understanding of what strategies "work well" ......and what strategies "work", but there are longer losing runs . 

I'm certainly not going to check out any more results, as the time it takes to do so thoroughly is
"considerable"!

It might be that you are only applying your strategy to certain race types......if that is the case, then the 200 fig might be a reasonable amount ( my check was on almost every race in Britain, at different points in season and on NH/AW and Turf.)

Hope you get some more replies from mathematicians....but your initial post is not comprehensive enough re what you are doing, so perhaps that is why you are not getting many replies on thread.

GL
By:
JLivermore
When: 14 Mar 12 12:35
Under new username  (posted as martinch).

Ideally we would have more info on odds and comm level to answer this question properly and my answer below  isn't perfect, but from what you say we can do rough approximation:
1 - number of trials = 212
2 - Assume all odds are 3.4, so p = 1/3.4*=0.294
3 - we would expect to win on 212 * 0.294 = 62.4 events given 1 and 2
4 - you actually won on 72.6 bets (I backed this out from 35 points profit at 3.4)

If you use excel you can use a formula
=1-BINOMDIST(72.6,212,0.294,TRUE)
1 - this gives the probability of winning in 72.6 OR MORE events GIVEN YOU HAVE NO EDGE = 6.5%
So it's quite unlikely that you are purely getting lucky and therefore it looks fairly convincing that you have an edge.

I would estimate your edge at 8% (we could have a huge debate on how to get 16% to a guess of your ROI on your next 212 bets).  When you should ramp up... that's really a person specific question - I would just suggest doing things gradually (and stakes should always be a % of bankroll).


And the more your average odds vary around 3.4 the more issues with the stuff above.
By:
Thin and Crispy
When: 14 Mar 12 12:57
...and meanwhile back in the real world....you can calculate until you're blue in the face but it don't mean a dam thing as far as predicting future returns.I have methods which will show good returns for 1500 bets and show a loss for the next 1500 bets.....the only way to find if a method works is by putting down the cash..... and if you're in profit it works!
By:
JLivermore
When: 14 Mar 12 13:16
isn't that an expensive way to find out that something doesn't work?
By:
Thin and Crispy
When: 14 Mar 12 13:45
Not if you're betting in 1p units
By:
JLivermore
When: 14 Mar 12 14:06
Grin

you'll get banned for talk like that
By:
roadrunner46
When: 15 Mar 12 14:13
ive tested my plan.
37.5 strike rate
272 % ROI
on 240 races]Laugh
By:
JLivermore
When: 15 Mar 12 15:13
if I give you £2.40 can you test it for me properly?
By:
roadrunner46
When: 15 Mar 12 16:07
if you cant read the form book.
never going to make it pay.
there are plans can help you.
putin 27/1 winner reading form
found that one.Laugh
By:
SHAPESHIFTER
When: 15 Mar 12 17:37
When should you increase it? Why not now?

Start with a hypothetical bank of £200.

Build conditions to bet 3% / 4% or 2.5%.  The 'conditions' are built around 'sub-groups' of your systems.  You start at 3%, notch up based on analysis.  Notch down if you have had weaker results in a certain type of event (i.e. if football, if your strike rate is low when the system brings up a certain team, pull back to 2.5%

Increase by the odds won / decrease on loss by the stake.

If you traded to reduce your risk, this doesn't get factored into your 'hypothetical bank' since it won't be an indicator on your performance.
By:
Cuban
When: 15 Mar 12 22:26
roadrunner46     15 Mar 12 15:07 
if you cant read the form book.
never going to make it pay.
there are plans can help you.
putin 27/1 winner reading form
found that one
.

I like this one. Reason I created a backing strategy in the first place was to break even with it and reduce my PC liability, it seems I have been too successful and created a profitable backing strategy which generates only 18% commission on profit - was hoping for 40%!

Still I bow to your superior knowledge roadrunner, must get my head in the form book as there is obviously no other way to make money, silly me.
By:
Cuban
When: 15 Mar 12 22:30
JLivermore - thanks for the maths, I dusted off my old A level books - going back many years and came up with the same conclusion. I have calculated a margin of error of -/+ 6.9%, which means worse case scenario it is just better than break even. However, this is all theoretical of course.
By:
roadrunner46
When: 15 Mar 12 22:43
so you make money laying horses.
how comes 215 bets you only made
ROI 16%. yet your strike rate was 38%
your bets werent level stakes?
By:
roadrunner46
When: 15 Mar 12 23:12
yes your sums add up.forgive me.
my ROI 272% how many times
is that more then your ROI of 16%?
im not good at maths theory.
By:
Chilly the Dog
When: 17 Mar 12 21:21
There are two main things to consider:

You need to verify your theoretical edge and...
You need to use a decent staking system, such as kelly (or a modern modification), to place optimal bets.

To do the first, you need thousands of bets to be fully confident.
For the second, you need to know the "true" price of what you are betting on and the price you're getting, from that you can work out what percentage of your bank to bet to optimally grow it over time.

In general, up your stakes slowly, don't get carried away if you end up with a sweet winning streak. Remember that if you're getting an edge of more than 5% you're probably wrong (it's not impossible, but very rare).
By:
bf_fananatic
When: 24 Mar 12 10:56
So much seems to be wrong in your post that it defies belief, for starters....

point(1)
average odds of 3.5 multiplied by strike rate (SR%) of  38% would yield an average ROI of 133%, ROI meaning Return on investment.Your figure of 16% ROI neither refers to the implied ROI figure of 133% nor to ROI generally as its more near to POI figure format/profile or profit on investment.

point 2
A sample size of 200 is extremely small and no clear indicator to any robust profitable system and 2000 sample size would be a much better hallmark to gain confidence in any system!

point 3
Your average odds of 3.5 seem quite small and must refer to favorites of which if you were to look at many 200 sample size segments you could quite easily fall upon 116% roi or even 200% roi along with terrible losing runes equating to say 50% ROI, furthermore favorites are extremely difficult to make a profit with because all the value has been drained out with pricing mechanisms like trading/stake amounts and bookmakers over-cutting industry prices via track side bookmaking monopolization.Most pro punters will be betting on average odds of between 5.0 and up to 11.0 and certainly not relying on unprofitable short odds.

point4
You have not mentioned if you have back fitted your profitable sample runners as this is the easiest mistake to make when looking for profit as what appears to be profitable in the past may and does not generally reflect as profitable into the future and has more to do with natural variation than revealing hidden wealth and will actually mostly plummet into loss in the near future to naturally offset the peak that you recently observed
By:
Glasgow Brian
When: 24 Mar 12 13:19
Can somebody please explain BACKFITTING ?


If my laying strategy consists of the following rules --- Have i backfitted in any way and if not can somebody give me an example ?  thanks


1.  Fav
2.  Won last time out
3.  Won by under 1/4 length
By:
bf_fananatic
When: 24 Mar 12 14:03
I shall try to explain a back fitted system and where it can lead to problems if I may;

When looking at past data it is possible to see many peaks and troughs in the data where strike rate and ROI are good and thus profitable but you cannot say that this data then is always profitable as is the data trend based on natural variation or on a robust bias with a theoretical edge that provides constant profit?

In your example you are using a system based on a theoretical edge that being you are sub-dividing all favorites on horse racing into ones that have won last time therefore more likely to have either a penalty, raise in race class or drop in starting price due to attracting stakes.
Then you are sub-dividing into horses that have won well and those that were not clear of the competition and less likely to repeat a win.

The system seems to have good theoretical sense as a possible profitable laying system and not one that was simply back fitted that looked good retrospectively with no theoretically biased profit.
By:
bf_fananatic
When: 24 Mar 12 14:46
p.s. your system for laying will make even more money if you look for a change of jockey since the last win and should give you as much as 5% gain in ROI%, this is because the factor that led the original jockey to make any correct deciosions in ridng the horse to its win is no longer present.Wink
By:
Glasgow Brian
When: 24 Mar 12 19:09
cheers bf - thanks for your reply Happy

Could you please give me an e.g of a backfitted system ?

thanks
By:
DFCIRONMAN
When: 24 Mar 12 21:15
From FLATSTATS :-


Why Systems Fail

Systems fail mostly because the core rules are all wrong to start with. A punter creates a system with rules which are not sound and then procedes to back fit the system to make the results look good.

Back fitting is like shooting an arrow at a tree and then painting a target around where the arrow lands. In other words you are making the system fit the results. What usually happens is a punter thinks of a basic idea, looks at the data, and whittles away at the data until the results look good. This is not a sound way of creating systems and will lead to the system failing sooner rather than later.

Examples of backfitted systems include rules such as:

    Only back the horse when priced between 2/1 and 8/1 - Why 2/1 and 8/1? Why not 3/1 and 9/1?
    Don't back 3yo or 9yo horses - Why 3yo and 9yo? Why not 4yo and 8yo?
    Best with field sizes of between 11 and 13 runners - Why 11 and 13? Why not 12 and 14?

If a rule can not be justified then it is probably best not to use it. Only if there is a specific reason why the rules should be used should it be included. An example of this could be: 'Only back colts and geldings'. If this rule was applied to an all weather system then it is justified as there is a distinct gender bias on the all weather.

Systems can fail if the punter does not hold his nerve and ride out any (inevitable) losing streak. This all boils down to money management. If the punter does not stake an optimum amount on each horse, and prefers to place amounts based on hunches, or luck then he is going to find his bank becoming volatile and ultimately end up depleted.

A whole new topic can be started on staking plans but suffice to say: if a punter does not understand that he needs to control his stakes when the heat is on then he may as well forget the system from the start. Overbetting after a win, underbetting after a loss, or missing out a few bets to save money is not a sound way of betting and will also ultimately deplete the bank.

Another reason why systems fail is that some of them are boring to operate. The strike rate and profitability maybe good but the number of contenders it produces is low. Each day punters would have to check the racecards manually to find a system bet. This could go on for ages and many punters will just give up on the system.

One way to get around is this is to us the Flatstats System Alerts feature. This takes all the hard work out of finding contenders. Create your system, set the 'Alert Me' option to on and each day the racecards are scanned and any system bets found are listed on your own horse / system alerts page.
By:
bf_fananatic
When: 24 Mar 12 21:19
It will be easier to explain with an example what exactly a back fitted system would be and it is more to do with how one finds a system rather than it having any particular style or format which could be anything really.

Lets use a roulette wheel as an example, Fred wanders past the roulette wheel at his local casino of which it is his first visit with a friend, he notices that a player always puts chips on four squares on the table and seems to be winning more chips than everyone else, in fact after half an hour he takes what seems to be a huge amount of chips off the table and cashes them in much to the high praise of friends and other players who delight in his sudden good fortune.

Fred decides that he will will go to the casino tomorrow night and put chips on the very same numbers as he reckons this is a good system from what he has seen and the fact the numbers came up again after the winning player had left.

What do you think happen to Fred when he returned with a sum of money in the hope of winning big using this new system that he has noticed worked well the night before ?
By:
bf_fananatic
When: 24 Mar 12 21:45
The whole subject of backfitting a system leads on to a very important point that will occur in most gamblers systems as for me you have 2 types of selections.

(1) profile selected

This method is probably the most widely used but can be the most problematic as it avails itself to the backfitting negative method of finding profit.Profile selected assumes that a certain set of condtions are present in the form of the runner or team and/or inthe race or match type as well as weather and seasonal conditionas which play there part too.

This method also can lead to much fewer selections as the more fussy you become about perfect data the less selections over the same time you end up with and some backers may only back once or twice a week using this method which cant be good in the long run to make good monthly units of profit.

(2) rating selected

This is the best and powerfullest method for finding winners as it does not reduce the amount of selections possible but puts all the runners or teams in a set order reletive to there merits which can be anything you wish to add in what ever amounts suit best using a points system.

The beauty of this method is that you cover all the runners or teams and have a broader or total spectrum of chances realtive to points that can be checked overtime to index realtive chances.

I wonder if anybody else has made this discovery as I have tried both types of system selection and know that raing is by far the powerfullest method possible and highly lucrative!
By:
bf_fananatic
When: 24 Mar 12 21:46
typo*rating
By:
DFCIRONMAN
When: 24 Mar 12 21:59
I have used RANKING order re RATED races ...then if odds are "value" to lay or back, depending on position in ranking ordr, bets should be placed.

EG ...If horse is ranked 3rd top rated and is 2-1 SP......there is no "value" backing or laying it.

If odds are say 3-1 SP....it is an automatic back.


If odds are say evens.....it is definitely not a back....and could be considered as a lay....but as in top 3 rated I probably would not lay it.

=============================================================

I am taking the RANK order as a strict guide to finding value re bets....

Of course I have had to BACK CHECK results on approximately 2,000 races against system ratings and odds to "see" the uses for the system. Some might believe this is BACK FITTING.......but is not IMO.....it is just arriving at a strategy that works due to the value in odds and the strength of system re rating races.
By:
bf_fananatic
When: 24 Mar 12 22:24
very intresting points DFC and quite true, I am compiling at the moment and not at the point of back checking(not back -fitting) the ranking and points aspect againt market price and position asI have not reached adequte sample size to find reliable trends.You have quite a volume I see and can afford this approach and its strange that I am curremtly noticing the same effects that you mention, If a favourite is far down the pecking order in terms of pointer/points then it fares much less well than when it is clear top-rated. I am convinced that in reverse one could find very good lays as some runners mark up very poorly and yet can even be odds on some times!

I agree that back-checking and back-fitting are 2 different methods,
sample volumes and thoerical edge being the deciding chalk marks !
By:
bf_fananatic
When: 24 Mar 12 22:30
p.s. When I refered to "not finding relible trends" my reference was in aspect of ranking order against price and not top rated which are proving quite profitable.
By:
bf_fananatic
When: 24 Mar 12 22:32
Recently for instance at the Chelenham festival I achieved about 130 points profit BSP with my top three rated for each race, not bad for a four meeting range of races!
By:
bf_fananatic
When: 24 Mar 12 22:42
Chelts Festival
my top 3 rated

avg BSP12.46
72.00
16.00
127.29
By:
bf_fananatic
When: 24 Mar 12 22:43
avg BSP 12.46
Sels    72.00
winners 16.00
BSP pft 127.29
By:
bf_fananatic
When: 24 Mar 12 22:45
8.53    Trifol    769.2728034    -1
12    Cinder    633.3841858    12
8.91    Steps     594.0085982    -1
12.17    Menora    978.2538647    -1
1.82    Sprint    977.123353    1.82
4.97    Al Fer    680.0416801    -1
90    Mon Mo    1812.208459    -1
7.8    Quanti    1391.394003    -1
7.25    Hold O    927.2809957    -1
1.79    Hurric    2389.644199    -1
12    Zarkan    1330.242344    -1
27    Overtu    1041.156383    -1
22.26    Gone T    880.117675    -1
19.07    Garde     864.0259387    -1
5.03    Scotsi    838.8437792    -1
1.61    Queveg    2431.737105    1.61
14    Shop D    718.2122933    -1
90    Dare T    701.4774868    -1
9.7    Triolo    598.6016709    -1
24.47    Educat    537.3360625    -1
9    Hunt B    459.7239613    9
6.6    Teafor    797.6351086    6.6
8    Allee     731.6793619    -1
    Soll42    618.8784105   
3.31    Simons    840.1837352    3.31
12    Cotton    702.3888444    -1
6.87    Monksl    673.0626609    -1
2.42    Grands    1737.413922    -1
31.42    Call T    640.4572072    -1
5.7    Bobs W    615.9766028    5.7
1.96    Sizing    2525.099322    -1
8.8    Big Ze    1807.29125    -1
5.9    Finian    1435.056067    5.9
34    Final     1108.520744    -1
8.74    Get Me    1056.558612    -1
0    0    893.0219852    -1
8.91    Kazlia    778.0535313    -1
36.43    Kapga     487.3386989    -1
55.32    Une Ar    474.0886466    55.32
17.73    Clonba    835.6933389    -1
16.5    Pique     787.9381496    -1
22.54    Champa    631.3158147    22.54
3.95    Peddle    1968.602944    -1
8.83    Crista    970.535644    -1
4.7    Sir De    854.6717242    4.7
19.28    Restle    764.5284285    -1
82.14    Reinde    581.7928427    -1
21.97    Pineau    510.3777877    -1
13.85    Albert    3004.542835    -1
7.43    Rubi L    2019.578899    -1
4.39    Rivers    1941.51851    4.39
1.87    Big Bu    1906.427506    1.87
5.5    Oscar     1789.737935    -1
11.5    Thousa    1640.162972    -1
11.84    Divers    1364.903682    -1
23.4    Giorgi    972.9666438    -1
9.04    Notus     940.5859334    -1
14.53    Becaus    2531.704456    -1
9.53    Sunnyh    1368.883744    9.53
81.06    Piraya    1229.202506    -1
9.74    Balder    1036.504591    -1
46    Countr    675.68133    46
6.37    Grumet    622.4553352    -1
    Clerk´    1154.313733   
24.39    Moon D    1124.326071    -1
10.5    Raya S    944.6095868    -1
2.5    Boston    906.8627315    -1
9    Brindi    857.5619839    9
18.1    Rocky     476.2006066    -1
357.89    The Mi    4375.997412    -1
4.5    Kauto     3414.765158    -1
2.86    Long R    3344.150816    -1
5.27    Cloudy    3141.163353    -1
13.5    Barber    996.641694    -1
20.72    Roulez    813.0969442    -1
17.18    Moloto    875.3076416    -1
    I´msin    875.1976888   
50    Rigidi    690.6798971    -1
26    French    960.6869577    -1
6.26    Kid Ca    781.8388778    -1
56.69    Idarah    690.1396637    -1
By:
Glasgow Brian
When: 24 Mar 12 23:00
Totals

Bets    Wins    Win Strike Rate    Win %Return at Estimated Betfair Odds
All Selections    922    101    11.0%    70.6%


Results broken down by Year

Bets    Wins    Win Strike Rate    Win %Return at Estimated Betfair Odds
2004    103    15    14.6%    84.3%
2005    125    15    12.0%    77.6%
2006    134    14    10.4%    72.3%
2007    143    15    10.5%    65.1%
2008    155    14    9.0%    57.2%
2009    137    15    10.9%    76.5%
2010    125    13    10.4%    67.0%






The above is a laying system i derived off the massey site  - AUG/SEP  - Flat

How likely is this to perform well in the future ?

cheers fellas .
Page 1 of 2  •  Previous 1 | 2 | Next
sort by:
Show
per page

Post your reply

Text Format: Table: Smilies:
Forum does not support HTML
Insert Photo
Cancel
‹ back to topics
www.betfair.com