Flat Handicap Weight Rank – Win A/E, Place A/E, Going and Distance
I have gone back through some historical Flat handicap work to look at weight rank in a bit more depth.
The analysis covers
81,604 Flat/AW handicaps and 805,452 runners. After removing nine races containing invalid SP values, the main win analysis contains
81,595 races and 805,360 runners. For the place analysis there were
79,415 valid races and 797,066 runners. Pasted text
The idea was not simply to look at raw strike rate. For every race I converted SP into implied probability and then
normalised the probabilities so that the race summed to 100%. This gives a market-derived expected win rate and therefore an
Actual/Expected (A/E) figure.
I have also now added expected place probabilities. As there are no historical place prices in this database, these were derived from the normalised win probabilities using a
Harville/Plackett-Luce model. The actual Place field in the database determines whether a horse placed. So the place A/E is:
Actual places ÷ expected places
The place probability calculations balanced correctly at race level, with effectively zero probability-sum error. Pasted text
Weight rank overall
The first thing that stands out is how orderly the basic weight-rank relationship is.
On Turf, top weights won
14.40% of their races, second weights
12.95%, third weights
11.86%, fourth weights
10.78%, and the strike rate continued falling as we moved down the weights.
However, after allowing for the market expectation, the Win A/E figures are remarkably close to 1.00:
| Weight Rank | Win % | Expected Win % | Win A/E | Place A/E |
|---|
| 1 | 14.40 | 14.22 | 1.01 | 0.98 |
| 2 | 12.95 | 12.86 | 1.01 | 0.99 |
| 3 | 11.86 | 11.85 | 1.00 | 0.98 |
| 4 | 10.78 | 10.76 | 1.00 | 1.01 |
| 5 | 9.56 | 9.58 | 1.00 | 1.01 |
| 6 | 8.46 | 8.46 | 1.00 | 1.01 |
| 7 | 7.34 | 7.34 | 1.00 | 1.02 |
| 8 | 6.30 | 6.45 | 0.98 | 1.02 |
| 9 | 5.58 | 5.72 | 0.97 | 1.04 |
| 10+ | 4.65 | 4.64 | 1.00 | 1.03 |
So the simple conclusion remains that
higher-weighted horses are stronger horses, but the win market knows that very well. Pasted text
One interesting wrinkle is that the top weight's
Place A/E is only 0.98, despite its Win A/E being 1.01. That suggests the top weight is not generally over-performing throughout the finishing positions to the same extent as its raw win strike rate might imply.
Going makes a much bigger difference
This was where the analysis became more interesting.
For
top weights on Turf:
| Going | Win % | Exp Win % | Win A/E | Place % | Exp Place % | Place A/E |
|---|
| Firm/GF | 16.36 | 15.83 | 1.03 | 37.06 | 37.02 | 1.00 |
| Good | 13.62 | 13.45 | 1.01 | 33.78 | 34.21 | 0.99 |
| Good/Soft | 13.23 | 13.18 | 1.00 | 32.48 | 33.73 | 0.96 |
| Soft/Heavy | 13.06 | 13.48 | 0.97 | 31.21 | 33.83 | 0.92 |
Pasted text
The place figures strengthen the soft-ground finding.
On
Firm/Good to Firm, top weights place almost exactly as often as the SP model expects: Place A/E
1.00.
On
Soft/Heavy, they not only win less often than expected — Win A/E
0.97 — but their Place A/E falls to
0.92.
That is important because it suggests this is not merely a question of top weights failing to turn a good run into a win. On testing ground they are, historically, also finishing in the places less often than the SP market would expect.
Distance and going together
Distance by itself did not produce a simple pattern. The more revealing results came when distance and going were combined.
The most interesting cells were:
| Distance / Going | Win A/E | Place A/E |
|---|
| 5f Firm/GF | 1.01 | 1.00 |
| 5f Soft/Heavy | 1.10 | 0.95 |
| 6f Firm/GF | 1.06 | 1.00 |
| 6f Soft/Heavy | 0.99 | 0.90 |
| 7f Firm/GF | 1.09 | 1.04 |
| 7f Soft/Heavy | 0.93 | 0.91 |
| 1m Firm/GF | 1.00 | 1.00 |
| 1m Soft/Heavy | 0.87 | 0.92 |
| 9–10f Firm/GF | 1.04 | 1.00 |
| 9–10f Soft/Heavy | 0.93 | 0.90 |
| 11–12f Good | 1.11 | 1.00 |
| 15f+ Soft/Heavy | 0.88 | 0.94 |
Pasted text
For me,
1m on Soft/Heavy is probably the clearest negative result so far.
There were
1,717 runners, with a Win A/E of
0.87. More importantly, the Place A/E was also only
0.92.
So this is not simply a case of top weights reaching the frame but failing to win. They were under-performing the SP-derived expectation on
both measures. Pasted text
The same general pattern is present at
7f Soft/Heavy:
Win A/E 0.93
Place A/E 0.91
and at
9–10f Soft/Heavy:
Win A/E 0.93
Place A/E 0.90. Pasted text
That makes the middle-distance/testing-ground area particularly interesting.
7f on faster ground is almost the opposite
One of the strongest positive combinations is
7f Firm/Good to Firm.
There were
3,128 top weights, of which 509 won:
Actual Win % 16.27
Expected Win % 14.98
Win A/E 1.09
The place figures were also positive:
Actual Place % 38.36
Expected Place % 36.74
Place A/E 1.04
Both the win and place measures are therefore pointing in the same direction. Pasted text
That is more persuasive to me than something such as
11–12f on Good, where the Win A/E is an impressive
1.11 but the Place A/E is exactly
1.00. In the latter case it may be more about converting competitive runs into wins than a broad performance advantage. Pasted text
A useful warning from 5f Soft/Heavy
The 5f result is also a good warning against making the theory too simplistic.
At 5f on Soft/Heavy:
Win A/E = 1.10
Place A/E = 0.95
So top weights actually won more often than their market expectation, despite placing slightly less often than expected. Pasted text
That alone tells me that a rule such as:
“big weights struggle on soft ground”
is too crude.
The interaction between
distance and going matters.
Recent results
There is also no evidence that the general Turf top-weight win relationship has disappeared recently.
For Turf top weights:
Last 90 days: Win A/E 0.99, Place A/E 1.00
Last 180 days: Win A/E 1.02, Place A/E 0.99
Last 365 days: Win A/E 1.02, Place A/E 0.97
Last 730 days: Win A/E 1.02, Place A/E 0.98
All history: Win A/E 1.01, Place A/E 0.98
Pasted text
So overall it remains a very efficiently priced part of the market.
What I take from it
My conclusion at this stage is that
weight rank by itself is not particularly useful as a betting angle.
There is a strong and very orderly relationship between weight rank and strike rate, but SP largely prices that relationship correctly.
What looks more interesting is the interaction between:
Weight Rank × Going × Distance
and possibly eventually:
Weight Rank × Going × Distance × Field Size × Horse Type/Size
The addition of Place A/E has made some of these results more convincing.
In particular,
7f–10f on Soft/Heavy ground appears worth further investigation because top weights are under-performing market expectation in both wins and places.
Conversely,
7f Firm/Good to Firm currently shows the opposite pattern, with both Win A/E and Place A/E above expectation.
This also makes the suggestion about using
historical stride length as a pre-race proxy for horse size/frame particularly interesting. Actual historical horse-height data is difficult to obtain, whereas average stride length from previous sectional runs may give us something measurable to test.
I don't think the evidence says simply that “high weights are good” or “high weights are bad”.
