-----

Betting statistics from the  2008-2009 season

This is the first season review that I have been able to post since 2006-07. Unfortunately my 2007-08 campaign ended prematurely when I had a serious acccident while riding my road bike. The accident occurred on Christmas eve 2007, and I was forced to spend the next six months in hospital and rehab.

I have placed on this site two Excel files.  Both are in Excel tab-delimited text format, so they should be readable by any version of Excel.  If any file launches the Excel Wizard, simply follow the instructions.

The first Excel file contains my personal Betfair results for the period November 2008-March 2009 inclusive. They represent a level stake win bet on every horse in the top two that was offered on Betfair at equal to or greater than the V-30 odds, with the stakes normalised to 1 point per matched bet.

A summary of the results obtained reads as follows:

Total bets matched             1367
Number of winners               143
Strike rate                      10.46%
Gross profit                      5.13
Average profit                    8.60
Percentage gross profit           0.38%
Average level matched            12.58
Average level winners             9.60
Sum profit before losers       1229.13
Betfair commission               60.23
Net profit                    -55.10

The spreadsheet shows a good profit for November 2009 followed by a series of roughly break-even months in December and January.

February and March show an alarming decline, so alarming in fact that I was forced into a complete re-think.

This was not just a case of blind panic. The rolling ARCHIE scores (read Steve Tilley's excellent article) were telling me that something had gone seriously awry. The fact that the accumulated losses were due for the most part to Betfair's commission at 4.9% was small consolation.

I came to the conclusion that the arbitrary cut, limiting bets to horses in the top-two rated, was doing more harm than good.  I was saved from total catastrophe by the fact that I always operate with a generous bank, and trigger a stop-loss option when the total of losses incurred approaches 50% of my bank.

As I've pointed out before, HOOF ratings are in essence a stats-based odds-line. It is a characteristic of most odds-lines that the value to be found in a race is not assigned to one stand-out value horse. It tends to be distributed amongst a number of potential candidates.  Here is an example from a typical  small race:  The V_30 column represents the value odds with a theoretical 30% advantage to the backer.  The V_PLUS  column represents exceptional value at 40% or more in favour of the backer.
 

Worcester:4.20 Handicap hurdle:8 rnrs
HORSE                  TR-PROB TR-ODDS   V-30  V-PLUS
Two Miles West           0.234   10-3    11-2    7-1    no match at 11/2 or better
Olmeto Collonges         0.126    7-1    11-1   14-1    no match at 11/1 or better
Mith Hill                0.125    7-1    11-1   14-1    no match at 11/1 or better
Bob Mountain             0.124   15-2    11-1   16-1    matched at 12/1 = lost
Chinara                  0.104    9-1    14-1   18-1    matched at 28/1 = won
Raslan                   0.098   10-1    14-1   20-1    matched at 14/1 = lost
Little Shilling          0.094   10-1    16-1   20-1    no match at 16/1 or better
Quattrocento             0.09    11-1    16-1   22-1    no match at 16/1 or better

However, old betting habits die hard, and I have always been reluctant to bet on more than one, or at the most two, in a given race. My re-think at the end of March involved a retreat from that strategy.

And so to the second Excel file. This shows the results for a 'bottom fishing' expedition that involved betting on every V-30 horse in all-aged handicaps regardless of rank order.  A breakdown of the results obtained reads as follows:

Total matched bets                        956
Winners                                    54
Expected winners                           30
Strike rate for matched bets                5.65%
Gross profit to 2 GBP                     702.50
Total winnings before losses to 2 GBP    2503.28
Average profit per win to 2 GBP            46.36
Total staked to 2 GBP                    1908.78
Percentage gross profit                    36.80%
Average level matched                      31.86
Average level winners                      24.18
Average level winners @ sp                 15.01
Betfair advantage over sp                   9.16
% Betfair advantage over sp                37.90%
Average level winners @ best board price   16.66
Betfair advantage over best board price     7.52
% Betfair advantage over best board price  31.09%
Betfair commission  to 2 GBP @ 4.9%       122.66
Betfair commission as percentage of gross  17.46%
Net profit to 2 GBP                       579.84
Number of races                           215
Average level stake liability per race      4.45
Average liability per race to 2 GBP         8.89
Winning race strike rate                   25.12%
Current ARCHIE score                       19.81

My stake was, of necessity, the minimum. I averaged around 4.5 matched bets per race. In some races I had up to nine matched bets running for me, but after a while, I got used to the idea of an average liability per race of 4.5. I came to the conclusion that it was no worse a position to be in than if I had laid a 7/2 shot.

The campaign lasted from 11th April until 24th May when the motherboard on my Mac failed and I was forced to call a halt. I'll let the results speak for themselves, but please note that they are subject to the following caveats:

1. LIQUIDITY The sums that I was placing were the minimum allowable on Betfair.
In many cases, it is highly unlikely that one could have obtained a full or
even a partial match if one had scaled up the stake. Unless the
betting is very strong, I doubt whether there is much scope for a thick
bet at the extreme back end of the market.

2. TIME OF YEAR It has always struck me how certain times of the year
favour a rash of shock results. April is a classic case given
frequent changes in the going, and additionally, the change from winter
to summer racing. Whether the results I obtained are repeatable at
other times of the year remains to be seen. May, though profitable,
was more difficult, with a noticeable decline in liquidity.
Incidentally, I think that the relative lack of liquidity during the
summer months when the flat is in full swing, makes it unlikely that a
summer campaign will ever prove to be worthwhile.

3. LUCK IN GETTING MATCHED I decided quite early on to put my bids in
only when the market conditions were optimal - preferably more than 25k
staked and an over-round not greater than 102%. This generally meant
getting on around an hour and a half before the start of the first race
at any given meeting. On those terms I made a number of great matches,
some of which were clearly the best that the market had to offer.

4. THE BETFAIR ADVANTAGE OVER SP It's worth noting that the gross
percentage profit is not far off the Betfair percentage advantage over
sp. At sp, I'd have broken even at best. In fact, given that many of
my successes were less than the required V-30 at sp, it's likely that
I'd have made a loss.

5. RACE TYPE In opting for all-aged handicaps I chose well. I doubt
if the results could be repeated in novice events.

6. ARE THE RESULTS REPEATABLE?  I have no idea.  Trying to look
back from 11th April in order to work out what might have have been
over the course of a whole season, is virtually impossible, since the
Betfair prices that were obtainable are unknown.  The only hard evidence
is the Betfair sp.  On that basis, there is evidence that seems to indicate
that the the season as a whole was, in fact, profitable.  However, since
the Betfair sp data are a relatively recent innovation, there are no plausible
results for seasons prior to 2008-09.

On the plus side, the results are not backfitted, the ARCHIE score is
reassuringly high, and the results do not include at least three winners
at huge prices that went in at the Aintree Grand National meeting.  In
addition, around half of the winners that were matched at the required
V-30 price on Betfair, were 'under the radar' in the sense that they did
not qualify either at sp or at best board prices.
 
 

PERFORMANCE OF THE RATINGS COMPARED WITH THE ON-COURSE BETTING MARKET

It is generally accepted that the betting market at the off is pretty efficient, in the sense that starting prices are the most accurate predictor of a horse's chance of winning.  That they do not  perfectly represent a horse's chance of winning is a reflection of the bookmakers' take out, or percentage, also known as the over-round.  Thus, 6-4 shots don't win exactly 40% of all races, but more like 37%.  It is also the case that the longer the odds, the greater the bookmakers' take-out.  Thus 20/1 shots should, in theory, have a strike-rate of  4.76%.  In practice, it is more like  3%.  In other words, the true odds about a 6/4 shot are around 13/8, and the true odds about a 20/1 shot are closer to 33/1.  This is why bookmakers prefer to see outsiders win.

Consider the table below.
 

ODDS RANGE         MARKET PROBABILITY         HOOF 'TRUE PROBABILITY'
                                                                                                                        2008-09
Odds-on             0.58                             0.66
EV to 11/8          0.43                             0.43
6/4 to 15/8         0.35                             0.35
2/1 to 11/4         0.27                             0.30
3/1 to 4/1          0.20                             0.20
9/2 to 6/1          0.14                             0.15
13/2 to 8/1         0.10                             0.12
9/1 to 12/1         0.07                             0.10
14/1 to 20/1        0.04                             0.06
22/1 to 40/1        0.02                             0.04
50/1 to 100/1       0.005                            0.02
> 100/1             0.0002                           0.005

The table tells us, inter alia, that at the front end of the market, the predictive power of  the starting price and the HOOF 'true probability' rating are virtually the same.  Thus horses in the betting market range of 6/4 to 15/8 have a win probability of 0.35, which is identical to that of horses in the same odds range assigned by the HOOF program.  To put this another way, horses in the market odds range 6/4 to 15/8 win 35% of their races.  This is identical to that of  horses in the same HOOF 'true probability' range. (Actually, they ought to win just over 37% of their races, but they don't.)

If one looks beyond the front end of the market, we see that in the case of outsiders, the HOOF 'true probability' figure is a better predictor of  a horse's true chance of winning than the market.  For example, in the range 14/1 to 20/1, the strike-rate for horses priced by the betting market between 14/1 and 20/1 is 4%.(probability of 0.04).  For the ratings it is 6% (probability of 0.06).   In the range 22/1 to 40/1,  the market scores 2% (probability of 0.02), while the ratings score 4% (probability of 0.04).

It is therefore unsurprising if the ratings are best at finding value among the outsiders rather than at the front end of the market.  The downside of this, as I have frequently pointed out, is a very low strike-rate, and the need for steady nerves and an enormous bank.
 
 
 
 
 

-----

[ Home Page ]

[ Introduction ]  [ Criteria for ranking top three ]