curious

It's also close to the beach and Pitty likes it there.

Might have been from a delusion in the NZTR business plan about optimal field size? This is from the old 2010-12 one: 2.5 Field Sizes Due to punter preference, NZTR must ensure strategies to encourage consistent field sizes of 12-14 starters.

So what you really want to play with is a distribution where the centre shifts left or right depending on confidence but the range doesn't shift. So that would vary from event to event? I think I'm getting the idea but at the same time scratching my head as to how you would do that. I think you'd need to set some rules which you also can play with around how much you shift the centre relative to some sort of shift in confidence score which presumably you already produce. I think that could be done.

Yeahh, if you chuck the outliers there's probably no significant difference between 7 and 14 going by that.

ok, so not centred on the mean. A kinda slightly skewed distribution but centred on what?

ok so you are maybe saying a non-normal distribution but still centred on the mean with the tail on one side longer than that on the other?

OK. I had to read this post a few times to grasp it but I think I'm getting the issue now. What you are thinking you might want to use is not a normal distribution centred on the mean but a distribution centred on the midpoint of the range (excluding outliers)?

Not so last I looked. Think 7 was the highest per runner. Not saying that is cost effective but 2 x 7 horse fields produce considerably higher turnover than a 14 horse field.

It's not just fitness either and I've done it myself here. But where you have a niggly difficult to ascertain soundness problem it's tempting to use races as a test of the current hypothesis and solution.

I quite like 6 - 7 horse fields as betting propositions and generally they have the highest turnover per runner.

This is another problem with trying to assess chance in NZ events. You can assess suitability but it is hard to assess fitness. In the UK and US you can assume it.

Yes. There are countless examples but a memorable one for me, and this is going back to the seventies, is Alleged. As a four year old he won the Royal Whip in the spring and didn't race again till autumn due to health issues. His first start back he broke the Longchamp 10f course record in the Prix du Prince d'Orange before going on to win his second Arc in his only other start. I really wish they would look at this here if they seriously want to increase attractiveness of the product to punters and increase wagering revenue.

Yeahh and that was exacerbated a few years back now when there was a push to start horses in races in preference to trials for that purpose to increase starter numbers. It was particularly so when we had free racing so it became more cost effective to "trial" horses on racedays rather than at trials and there was a concurrent significant increase in trial fees. I said at the time it was folly because it would reduce punter confidence if they could not know whether horses were there to try and win or were just racing for conditioning purposes. I think that move is yet another thing that has come back to bite them on the bum in terms of revenue. I.e., pursuit of a target of starter numbers to increase revenue, failing to understand the situation from a wagering perspective. It was based on falsely interpreted data anyway in regard to the relationship between starter numbers and turnover. They saw higher turnover in races with higher number of starters but ignored the fact that turnover per runner was higher in races with less starters (at least 7+). Plain stupid.

Maybe "its merits" allow for its current state of fitness?

That makes sense and if I ever figure NZ racing out to a point where I'm comfortable betting on it, I may have to do that. I should say that while I don't assess tempo or sectionals, I do assess settling position though in a historic way for individual runners, not in a predictive way for today's event as in a speed map. In that respect, you could say that I do assess opening sectionals in a way. Again however, that data is included in standard form guides in the US (time and position after 400m). I don't know if it is for NZ without doing a lot of work?

I'm curious Mardi why you would suspect that you might be better to use a non-normalised distribution. Is your time assessment data not normally distributed?

I'm pretty sure that is against the rules of racing here also though it seems to be turned a blind eye to. To the point where trainers can publicly announce things like that a horse will "need the run".

FTF, wrt time. I differ a bit from barryb a bit on this as I only use whole race time to assess ability and I use lifetime starts for that. I do not try to assess pace/tempo, use sectionals for anything etc. Not that there's anything wrong with that and some do that successfully. It may be that it is more important in NZ racing conditions. In the US the fastest sectional is usually the first whereas here it is usually the last. To use whole race times as an assessment of ability, the critical factor is to establish par times for different courses. This is much easier in the US where one track may race say 100 days a year, cf here where they may race 1 or 2 days. Somehow you have to be able to compare a time on one track and track condition with another. That requires establishing some sort of par times. I don't think raw times are much use at all. They need to be adjusted for both ability of the runner/s and unknown variables like whether the distance is accurate, track condition etc. The Beyer speed ratings in the US were based on par times of track records. This is a reasonable proxy when you have several hundred events a year on a single track. Completely hopeless here so you need another way of establishing those. You may be further ahead than I on this. I've spent 18-24 months on and off trying to get a handle on this for Oz racing and only now am confident enough to probably start betting on Sydney Melbourne metro events next year though still a few wrinkles to work out. Still a bit at sea elsewhere in OZ. There's quite a bit been written about establishing par times and a variety of ways of doing this. Usually this is a range (sometimes one standard deviation for example). Not sure what works best for NZ but do let me know if you find the answer to that FTF

Thanks FTF. As a p.s., I meant to add that I wouldn't use raw strike rate to assess this (or anything for that matter). I think you need to use strike rate cf. expected strike rate. Mardi and I spent weeks trying to convince the powers that be at NZTR that you can't do this when we did some work for them assessing the impact of the changes in the handicap system a few years ago. Starting prices were not accurate enough to use for expected chance, though we found that actual handicap ratings were a reasonable proxy for that even though they were skewed in favour of females and higher rated horses. Problem was we didn't have that option for maiden races. I'll come back to you on your time related questions when I get a chance.

Outside of starting with-in 21 days the SR % drops away significantly. That's interesting and a bit surprising. You can probably have a look at that in your data set FTF. I assume that's for NZ? Maybe Oz as well? I don't find that in the US, though there is a drop-off for less than 15 days. UK Mardi?

Just tried that and it works but don't think the quality is as good as watch&bet?

Is it down or is it just me?

Must be a couple of winners in there barryb. Best of luck.

Because I thought it was interesting, and some have indicated that they are having a go at framing their own markets, I'm going to reactivate this thread with a few questions for the gurus on here and hopefully the thread can stay constructively on track. Barryb and Mardi in particular, and anyone else with expertise and knowledge in this area, 1. When assessing a race or runner, how many variables do you consider? 2. How much time on average do you spend assessing a single race or maybe 8 race meeting say (excluding any reassessment due to change of track condition, scratchings etc.)? 3. I know that Mardi uses apps for simulations to get his estimate of chance and maybe to assess his expected times. So, my question is what proportion of the assessment is software based and how much is done manually? 4. Mardi, I think you said somewhere that closing BF prices are a pretty good approximation of an accurate market. I'm also pretty sure I read a study (I think done in Hong Kong) showing that the correlation between closing tote prices and results was only about 35-40%. My question is, on events where there is sufficient volume for it to apply, how much better is a similar correlation with BF closing prices?

H10 at puni as well.