Niall, I'll chime in my two cents here regarding figure making. Not sure if this will help or not but here goes.

One way to make figures is to come up with a set of pars for different race levels. You then take a look at the times the horses ran for the day and calculate the figures for all of the winners and compare what those raw numbers look like to your par charts to try to measure how much the speed of the track affected final times. You want to come up with a track variant for the day. So using Beyer numbers, assuming you've got 5 races all run one turn on the dirt, lets say you get this:

Race 1 Par 80, Fig 87 = 7pts fast
Race 2 Par 65, Fig 75 = 10pts fast
Race 3 Par 80, Fig 84 = 4pts fast
Race 4 Par 75, Fig 82 = 7pts fast
Race 5 Par 90, Fig 93 = 3pts fast
Average = about 6pts fast

So your variant in this case would be 6pts fast. That's the speed of the race track that day. So you would adjust all of those winning figures down by 6 pts, giving those horses an 81, 69, 78, 76, 87 respectively. And you would then give the figures to the horses who finished behind the winners their associated figures based on the number of beaten lengths they were behind the winner.

Some of the inherent problems with doing it that way (among others but to keep it simple) is you are assuming that the track speed didn't change at all during the day, you are also assuming that the field was an average field for that racing level, you are assuming the teletimer was accurate for each race, and you are letting each race potentially impact the figures earned in the races before and after it. Those are very big assumptions that often turn out to be wrong.

Another problem is that you often won't have enough data to support a variant (you may only have 1 race at two turns on the dirt that day and everything else is run 1 turn on the dirt or on the turf).

Another problem is that very often you will find days where your variants would look like this, +10, +9, +17, -4, +8. So if you average those you get 8pts fast. But when you apply the variant of +8 to all the winners and then the horses that ran behind them, in that 3rd race it turns out that every single horse in that race now gets a figure that is about 10pts faster than they have ever run before. Is that very likely?

So instead, what most figure makers have realized is that using par charts is a bad way to go. So instead they go off of the horses. What I mean by that is if you have 3 horses in a one turn dirt race who last ran 85, 80, 80, and your winning figure for today's race based off the raw time gives them 95, 90, 88, (all 8-10 pts fast) you would make an assumption that if you adjust each horse's figure down by about 9 pts that probably was the variant for [u]that race[/u] that day because its very unlikely that all 3 of those horses improved by 8-10 pts at the same time. And by adjusting them down by 9 Points you get figures of 86, 81, 79 for each - which would seem to line up pretty well with what they have run before and make sense. So in this case you are using all of the horses in the race to come up with the figure for that race to see if it makes sense. So what you do is you do the entire day, assign the figures for each race, and then take a look at all of your data to see if it all seems to make sense. You aren't necessarily concerned if the variants for each race seem to be similar because you aren't coupling the races together. You are going off the horses established form. But its even better when they all seem to line up because it can give you an ever higher degree of confidence that your figures are correct. But even if not, lets say you have that one race that seems to make sense internally to the horses who were in it, but does not make sense externally to other races that were run that day, you probably leave your figures as they are and schedule the race for review at a later time by flagging those horses and seeing what they run next time out.

So what Beyer and assoc. are saying is they have established figures for some of the horses that ran in the Mendelssohn race, and if they use those figures to help figure out the variant for that race they come up with a beyer fig they are pretty confident in because by assigning him that figure the figures for the other horses who ran behind him seem to line up, and that two people used that same method independently and arrived at the same figure.

You can still have some problems no matter which way you do it because you either wind up with bad assumptions or some "art." But going off the horses is the far better way to do it IMO.

When a horse freaks and wins by the length of the stretch like Mendelssohn did, I mean he ran off the screen first time on dirt, you don't have anything to really go by in his own pp other than what he has run on turf and synthetic. You've got the teletimer and the horses who ran behind him that you have figures for. He massacred all of them and it's probably wrong to assume they all bounced and ran OFF races. So you have to assume a few of them still ran their race. And if you do that you have to give Mendelssohn close to a negative 5. So come Derby day, we are going to be looking at a TG sheet where the UAE Derby winner is probably the fastest horse in the race by a fair margin, and this is a fast group of 3 year olds this year. If he repeats that race he is probably gone, but his sheet will scream bounce.

This game is never, ever easy.