First question this month comes from Dean Gardner from Melbourne, who writes: I'm a young stud who's only just got into racing, mainly due to attending Oaks Day at Flemington last year and finding it's the place to be in Melbourne during spring. After going to the races a few times since, along with frequent visits to my local TAB, I got to know one old regular who goes by the adage of "winning-form is good form".

After reading somewhere that last-start winners win about 30 per cent of races, my question is: Is this a truism, is winning-form good-form, or just another racing fable that doesn't stand up to the test?

Practical Punting Monthly readers may well remember a (true) story I wrote some time ago about Dino, a racing character I haven't seen around the tracks for some time.

However, I believe he had a big win on Lotto a while ago and bought a car with some of his winnings so maybe his interests are elsewhere nowadays - but I suspect he will be frequenting the racetracks once again during the next few months.

The reason I state this is because Dino is a picket fence bettor; he just loves those good horses with a string of current wins in their formlines.

So how do last-start winners, the "winning form is good form" horses, perform? The raw statistics would suggest not too good, at least not until a horse has strung together a run of wins.

The following table is based on races run on metropolitan tracks over the past twelve months.

20031010

Note: P/L7% equals the relative Profit or Loss on Turnover; LLR equals Longest Losing Run. 

Clearly, last-start winners do not perform as well as what many punters think, with a low strike rate of nearly 14 per cent, a poor return and a likelihood of long losing runs.

Obviously, there is a need to combine this factor with other factors if you want to make last-start winners the basis of any method that can go close to being a reasonable selection approach.

Horses who have won their last two starts have a higher strike rate but a worse return and still are likely to have long losing runs.

Horses with three wins on end perform better, but still not to an acceptable level and it is not until a horse starts to have that picket fence of four wins on end that they achieve positive figures.

The issue with the latter is that there aren't too many of them and the strike rate is nothing out of the ordinary, albeit they do make a nice return.

I could find only eight occurrences during the test period on metropolitan tracks of horses having won their last five races.

Unfortunately, both sets of statistics for horses having won their last four or five races are distorted by the win of  one horse, Blevvo, at Ascot on December 7, 2002. He started at odds of 25/1, paying $19.70 on TAB Ltd - a somewhat surprising dividend for a horse who had won his last five races. This win took his record to seven wins from his previous eight starts.

Just as surprising is the fact that during this winning run, Blevvo had started at odds of 25/1, 8/1, 20/1, 9/4, 9/4 and 4/1!

Unlike Dino, there are many punters who think that picket-fence horses are due for a loss so they look for other runners to knock them off, thus causing those good dividends.

Certainly, some back-to-back winners pay extremely well on occasions.

There is a belief amongst some smart operators that mares in winning form tend to hold their form better than male horses.

This is a very creditable opinion and one only has to think back to the great mare Sunline to get some appreciation of this factor; she strung together eight wins in a row on one occasion, five wins in a row on two other occasions and a sequence of four wins in a row on yet another occasion.

But there are other fillies and mares with less credentials and of more recent vintage who have also strung together a picket fence of wins: the then 3yo filly (now 4yo mare) Bollinger did so during the Sydney autumn carnival, winning four in a row, including the Group 2 Surround Stakes and the Group 1 Coolmore Classic.

Makybe Diva, a mare trained by David Hall, strung together six wins in a row last spring, going from winning a Maiden at Wangaratta on a heavy track to winning the Group 2 Queen Elizabeth Stakes at Flemington (at odds of 11/2) on a good track, in the one preparation.

But male horses are also capable of stringing wins together: Lonhro already has won eight times in a row on one occasion and five times on another, while the now-retired Australian Horse of the Year, Northerly, had winning runs of five and four during his career.

On August 2 at Cheltenham Park, Roman Arch and Full Spec both added wins to their picket-fence performances, while on the same day at Doomben three horses, Star Of Sequalo, Promising Joy and Pekan Lady, achieved similar feats.

At the time of writing, 3yo filly Star Of Sequalo has strung together five wins in a row, winning back-to-back as a 2yo in Sydney and adding three more successes as a 2yo and 3yo in Brisbane.

My advice is to concentrate on horses attempting at least their third win in a row and only consider having a bet if the horse is at a value price, which should be defined as being not odds-on, not rising sharply in the weights and with its price relative to the strength of the field at this meeting.

Russ Hills contacted me by email to ask for some advice about improving his rating processes: I discovered your website and must say that I'm impressed by the performance of your published ratings. The background to my question is that I have been doing my own ratings for a few years now, basically following weight ratings (Scott) methodology.

I use the Wizard's detailed ratings for each runner as the base (usually, best of last 3 runs, but not always), I calculate the rest with a computer, basically following the book.

About a year ago I decided to further analyse the performance of my own ratings and to my surprise I found that the performance of my top-rated win percentage varied considerably.

For example, for 1 000m races, it was 31 per cent, whereas for Moonee Valley 2040m races it was only 8 per cent,  with the strike rates for the other distances somewhere in between.

As a consequence of discovering this I don't bother rating Moonee Valley 2040m races any more, while I don't bother with any distances where the top-rated has a strike rate of less than 20 per cent.

While my overall results have improved since adopting this approach, my results tend to be flaky, depending on the track, e.g. for 1200m races my strike rates are:

TRACK1200m1600m
Caulfield18%5%
Flemington0%32%
Moonee Valley13%20%
Sandown8%25%

My questions are: What is your opinion on this and how can I go about identifying the factors letting me down at those distances where the strike rate is under 20 per cent?

Why are the better strike rates happening at some tracks and distances, yet not others?

The approach of playing to your strengths is a sound one. However, the low strike rates at some distances/ tracks are surprisingly low, e.g. 1200m races at Flemington, Moonee Valley and Sandown.

Have you ever attempted some post-race analysis on these selections?

Why didn't you have the winners rated on top, or conversely why did your top-rated selections fail?

There may be a pattern that could assist in improving your strike rates.

Where your top-rated selections performed badly, how did your second or third top-rated selections perform?

If your second or third top-rated selections outperformed your top-rated selections, then there is something not quite right with your ratings methodology, as the strike rates should be more or less linear in performance.

If, on the other hand, they also performed poorly, then there is something definitely wrong and you may need to go back to square one and re-assess your total approach.

My own ratings approach is somewhat different. It takes a contrarian approach to the Scott methodology in regard to weight by actually rewarding horses carrying the higher weights (but that's another story).

However, a few years ago I did take a more traditional approach to handicapping and I found that the best method was to use not one rating but a sliding scale approach of the last three runs (assuming all three runs are in this preparation) of 50 per cent of the rating of a horse's last run, 30 per cent of the rating of a horse's second-last run and 20 per cent of a horse's third-last run, i.e. taking the Wizard ratings for Converge, the topweight in race 6 at Sandown on August 9, of 57.5, 55.0 and 52.0, I would have derived a rating of 55.4. Where a horse has had only two runs this preparation, I use a 65/35 split.

Using that approach increased my strike rates notably - 28 per cent for the top-rated, 44 per cent for the top two, and just under 60 per cent for the top four. 

Another approach that may assist you is race-mapping, which may alert you to the fact that your top-rated selection may not be suited by the likely pace in the race or its position in the run.

As you would be aware, the best horse, which you may have as your top-rated selection, does not always win.

More than any other factor, this occurs because it suffers bad luck in running, which may be the result of a bad ride, being trapped wide or being blocked for a run. What are the chances of your top-rated selection suffering such a fate?

Finally, analyse the performance of your selections on wet and dry tracks to see if there are any notable differences in performance: my own experience has been that my ratings perform better on firm going.

By E. J. Minnis

PRACTICAL PUNTING - OCTOBER 2003