Graphical
Analysis
of Punting Strategies
using BetGrapher
In this article I hope to show you how I use BetGrapher
to analyse punting strategies graphically. There is an old
proverb "A picture is worth a thousand words",
and in the analysis of punting strategies this is very true.
There are
a lot of statistical tests that you can run when trying to
analyse your punting and I have run many of them myself.
However the most common means of analysis I use these days
is simple graphical analysis. BetGrapher is a program which
I use extensively to help me analyse and manage my punting
activities. I find that I can often tell at a glance whether
a strategy I have been following is performing as expected,
and if not, I can take remedial action.
Case Study
The case study I am going to work is a current strategy
which I have been paper trading for a few months. I developed
this strategy back in February / March this year on data
to the end of January. The base performance was 1178 bets,
S/R = 29.5% and POT = 25.5%. On the face of it, a pretty
reasonable strategy, and one that I would hope could perform
to that level. Here is the graph of the performance over
that time, with the blue line representing the profit.
My initial point of concern was the period between bet number
600 and bet number 900. This was 300 bets where the strategy
was going through a poor run and losing. Clearly the peak
at bet number 600 was due to a long priced winner, however
studying
the graph it can be seen that if this horse had not won,
the graph was still increasing nicely before bet 600 and
after bet 700, so the strategy still appears to be profitable
without this winner.
Rolling Bet Sequences
The sequence of 300 bets between 600
and 900 produced a loss,
so I decided
to have a look at every sequence of 300 bets over the entire
data period. This "rolling sequence of bets" is one of
the very powerful features of BetGrapher. It allows you to
monitor every single
sequence of an adjustable number of bets, in this case 300
bets. This means that it works out the S/R and POT for the
300
bets
between
bet
1 and bet 300, then the sequence from bet 2 to bet 301, then
bet 3 to bet 302, and so on. It then graphs these sequences.
The reason I look at rolling bet sequences is that where
you enter a sequence is very important to the initial performance
of your betting. Someone who entered the above sequence at
bet number 600, at the very peak of the run, would have to
wait for over 300 bets before they were into profit on the
strategy. On the other hand, someone who entered at bet 800,
at the low point of the run, would have been in profit immediately.
The same strategy, yet two completely different results.
The graph of the rolling 300 bets is shown below.
The two
horizontal lines represent the long term S/R (red) and
long term POT (blue).
The effect that the longshot at about bet 570 has on the
strategy is clearly visible from this graph, and this graph
also highlights another longshot winner at around bet 970
- one that I had originally missed when looking at the
graph of the profit.
The significant points about this graph are that in the
early stages, for the first period of the strategy, each
sequence of 300 bets was only producing around 10% POT,
and as low as 5% POT. So it is highly likely that you could
commence betting this strategy, and 300 bets later could
only be doing somewhere around the 5% to 10% POT. Another
significant point of interest is that there were very few
sequences of 300 bets that produced a loss, and this loss
was quite small at worst around -6% POT.
Without the two longshot winners, I would expect this
strategy to perform around the 10% to 15% POT range, and
I consider this to be worthwhile with the large number
of bets this strategy produces.
Rolling Time Sequences
Another useful way to look at past history is to consider
rolling day sequences. This is similar to the rolling bet
sequence, except that it graphs the S/R and POT over every
sequence of X days. Since I like my strategies to be profitable
over every 3 month sequence, I generally graph strategies
over rolling 90 day sequences. The graph of these are below.
This graph provides me with a few more reasons to be hesitant
about this strategy. There are many periods where the strategy
is in a loss for 3 months in a row. Since I often review
my punting in 3 month chunks this strategy would be coming
under regular review. Using the BetGrapher software I had
to increase my time period to 6 months before there were
any sequences where there were no losses. While I am willing
to bet long term profitable strategies over 6 months while
they lose, I doubt that I would do this for new strategies.
While this strategy may (or may not) be profitable, I am
wondering whether it would be bettable.
To bet or not, that is the
question
Given that there were a few negatives about this strategy,
I put it on my paper trade list, where I review it's performance
regularly but don't actually bet the selections. This policy
has saved me many thousands of dollars over the years by
avoiding those strategies which simply never work going
forward, regardless of how good or consistent they look
over past data.
Reviewing the strategy six
months on
Well it's now early September and I have been paper trading
this strategy for some time. The performance of the strategy
over that period of time is 319 bets for a S/R of 21.6%
and a POT of -8.8%. While it is currently in a loss situation,
the strategy has been in profits for periods of the paper
trading, and so the question is whether the strategy is
likely to return to profits or not.
The graph of the entire performance to the end of August
is shown below.
The downturn that has occurred since beginning the paper
trading is clearly visible, but in some respects is similar
to that of the downturn from bet 600 to bet 800. In fact,
at around bet 1400 it looked as though the strategy was
recovering and would emulate the recovery experienced around
bet 900, but the graph has since gone south again.
The rolling bet sequences
The real clue to the performance of this strategy comes
from the rolling bet sequences. I have had 319 "paper"
bets on this strategy, so I compared the actual performance
of my "live" 319 bets to all the previous sequences of
319 bets. I set the reference S/R (flat red line) to my
actual S/R of 21.6% and the reference POT (flat blue line)
to the actual POT of -8.8%.
This graph tells the story better than any test I know.
The actual S/R is clearly much lower than what it has been
for every single sequence of 319 bets in the past. The
moving red line only approaches the flat red line near
the end, which are the actual bets. The same story applies
to the POT. At no time previously was there a sequence
of 319 bets which produced a loss. There was one point
which basically broke even, but the actual performance
is -8.8%. At one stage during live betting the 319 bet
sequence actually produced a -20% loss which was significantly
lower than any previous result. (Yes, this includes some
of the pre-live bets but is still worth considering.)
Conclusion
It is quite obvious that this strategy is not performing
anywhere near it's previous performance. I have stopped
paper trading this performance, but have put it on a review
list where I will review it's performance after 12 months.
Some of the ideas in this strategy are good ideas, and
I may pull the strategy apart and rework it in a few months
time. It is possible that one or two "tweaks" will turn
the strategy around, but I don't hold out much hope.
Regards,
Chris Stebbing.
BetGrapher is available
for purchase at an introductory price of AUD$99.00.
Please click here to
purchase BetGrapher. A Demo version
containing the data used in this tutorial is
available
for download
from
my
forum. Click
here to go to the forum page. |