The difference is no one said it was going to happen anytime soon…
They are working on it!
1 Like
Back on the topic of what restricted races (pen splits etc) are actually for…
To me, it seems that the only real purpose of splitting riders (in most normal racing) is to enable weaker riders to compete for the win or at least a high placing. Ok there are a number of cases where we might want restricted classes without regard to strength, such as on the basis of gender, nationalities, age groups…but for the most part, it’s basically on the grounds of strength.
I base my approach primarily on the premise that once you’ve accepted that the main, perhaps sole. purpose of restricted categories is to give weaker riders a chance to win or place highly, then there is no reasonable justification for only giving this opportunity to an arbitrarily selected subset of these weaker riders. The idea that “really strong B cat” riders should get to win day after day, and “weak B cat” should get dropped off the start each ride, is complete nonsense - if the “really strong B cat” were genuinely strong, they wouldn’t be B at all. They are only “really strong” in zwift world because zwift chooses to match them up against weaker riders, which is an arbitrary choice of game designers that has no justification other than the laziness of the coders.
If what you really care about is that all riders will have some chances to compete at the front, a points system is trivial to implement and utterly foolproof. For example, if you want to ensure all riders to have some top-5 finishes, give points to the top 5 (only). If you want all riders to win occasionally, just give points to the winner. Etc. This is guaranteed to work for all riders regardless of strength, you only need to tune the boundaries between pens and the lifetime of the points in order that we don’t all end up as a huge bunch in the top pen or the bottom pen. This is not difficult at all, a decent programmer could do it in a day. The races might often contain quite a range of abilities, but each rider would have some races where they were at the strong end, and others where they were at the weak end (other than the very strongest and weakest riders on the entire platform, of course - and there’s nothing that can be done about them).
I actually don’t think it is important, or even particularly desirable, that all races are closely matched. As others have said, too close a matching just means a group ride and bunch sprint which would be boring to repeat time after time. I don’t think anyone reasonable would object to being well beaten on occasion, so long as it wasn’t every occasion. I therefore think that trying to match riders carefully is probably the wrong approach in principle regardless of whether it is even possible. I don’t want a bunch of riders like me every single race. I want to have the chance of dropping them one day, and a struggle to hold on to the pack on other days.
4 Likes
I think we are both describing the exact same desired outcome, but with different ways to achieve it (and maybe a slightly different view on what is easiest to do first).
1 Like
There are multiple ways that a group of riders will perform in a race. For example, the power curves of each rider, their endurance, and skill and current fitness; as applied to a specific venue that has different ways that challenge each rider. For example, flat, short climbs, long climbs, steep short, length of race, etc, again as modified by what the other people in the race do at in that race.
This quickly becomes a multivariate problem that simply cannot be computed to an exact result.
I’ll suggest that the goal is to create a group of people:
- where no single rider will win more than W% of the time where is W is > about 70%.
- no rider has more than L% chance of losing where L% is also about 95%.
In other words, no guarantee that you’ll ever win, but also no certainty that you’ll never win.
For these purposes win can also mean a top 3 or top 5 finish etc. And lose can mean dropped and finished in the last or second last groupetto etc.
For people that exceed these boundaries, move them up or down if possible. And that can be done by tracking via a ranking system that tracks peoples success at racing.
If we don’t want a ranking system, you could use a metric as simple as finishing in the top or bottom 5% of your last 10 races.
1 Like
Is it not what we call sport?
1 Like
That’s the conundrum. If the AI gets good enough, we’ll know the winner before the race starts and then we won’t need any pens or anti-sandbagging.
Excellent post.
I think that you really touched on an important point. The “strong C” (or D or B) is only strong in this virtual world and with this Cat system.
IRL, even in a local MTB race or amateur road race most of this riders would really have a notion of what a strong rider really means, and what is competitive cycling.
And i say it being a middle level Cat B (low Cat A in best shape) and getting totally destroyed in the type of local events that i mentioned.
So I’m not saying this to be cocky, it’s because I know that even when I’m in the threshold of jumping to Cat A, I´m still a normal rider. IRL I would not see the front of a local race … that´s reality.
So i agree with you. If the software is using an artificial way to improve the experience of the “weaker” ones (me included), the ideal system would achieve what you said; putting people in some races were they could be in front and even do a Top5 … in others they would be giving 100% to hang on to the draft or little attacks, or even getting dropped.
A system that give us variety of efforts, finishing positions; were we can be the hammer or the nail, would be the most exciting one.
Always getting beat like a dog it´s not engaging.
2 Likes
whoa… that’s weird
But it uses machine learning (a form of AI), and the more races that happen, the more accurate it gets.
AI really is like magic. It seems like it is constructing some sort of a ‘ranking’ of riders measured against one another based on ‘previous results’… 
8 Likes
…I don’t think that was the point. The point is that whoever is ‘strong’ or ‘weak’ is based on how we split the field, not some natural order of things. If you split the B-category in the middle, then the subsequent winners become the strongest Bs but ‘strong Bs’ become less strong relative to the tail end, without anyone’s actual performance changing.
3 Likes
Yes exactly. This is where pens, and how they are split, become important. They dictate the liklihood of a rider winning.
There’s no getting away from the fact that there can only be 1 winner (or 4, or 6 - depending on the number of pens).
If you want to mix up the likelihood of a specific rider winning, you have to be able to vary how the pens are split, they have to be dynamic not static. That could be determined automatically (evely split fields or something) or my preference, determined by the race organiser.
The author of the AI piece begins with two questions:
- How well will I do today?
- Who are the strongest riders in this race?
Today they are both incredibly easy to answer. 1. Depends on whether you are a strong B or a weak B
2. The top Bs (probably A FTPs but still in B on ZP)
1 Like
But remember that splitting the riders differently still only works if the riders are ranked in the correct order!
If stronger rider X is persistently ranked below weaker rider Y (by whatever metric you are using to sort the riders into pens) then Y will never win and X will never lose, because regardless of where you put the pen boundaries, X will always be able to enter the same pen as Y.
This is one of the critical flaws of 20 min W/kg. It simply doesn’t order the riders correctly, and contains no mechanism for correcting such errors.
With a points system it is not necessary to have the complication of dynamic pen boundaries. Rather, it is the riders that move dynamically between pens depending on whether they have won (placed highly) recently. If you can win in C but not in B, you will win in C until you get over the threshold to move into B, then race in B until your points expire, and you move down again. You’ll get some races where you are at the strong end, and others where you are at the weaker end.
Of course, it’s possible to change the thresholds between pens which might sometimes be useful to equalise field sizes. But riders will get a range of experiences and challenges regardless.
1 Like
Exactly this. Dynamic pens means a rider never knows against whom they are competing until the last minute (as they may get shifted to a different pen if an organiser is trying to spread out riders by entry numbers, rather than ability).
If we could just have the ability - as we do now - to set our own parameters and have them enforced, it would very simple to split pens by ranking. Let the organiser decide definitive boundaries, rather than have some algorithm decide that a 3W/Kg rider must race against 4W/Kg monsters purely because it wants to spread out the numbers.
2 Likes
You are actually agreeing with my point and not James’.
By dynamic I do not mean automated, I mean organiser defined. I think splitting the pens last minute based on the signups is too complex.
If you give the race organiser the flexibility, you do away with the issues related to static category boundaries.
A points system is far more complex than having a basic toolkit for race organisers where they can choose how the pens are split. As I have said numerous times before, once rankings come (which is a better approach than a points system, although we may be talking around the same thing) then this should also be an option to split pens.
If race organisers have different ways to split the pens, it solves for pretty much everything. No-one can complain - if you don’t like a particular race format, enter a different one.
If you limit yourself to a points system, that system has to account for a wide range of situations and complexities.
For example (this is off the top of my head, but I am sure we could come up with many more situations that need designing for):
If you have little or no points, that could be because you are not very good, or because you are new or have not ridden for a long time. If you are not very strong (today’s cat D) won’t it get frustrating when most the races you enter are won by strong riders who just happen to have no points yet? New riders or expired points. You never get the chance to gain points because there is always a rider or two that is racing well below their physiological capability.
It can all be solved for, but a simple toolkit for organisers to define the pen boundaries by whatever methods they see fit, that can have additional options and inputs added over time, resolves 90%+ of today’s problems with racing.
1 Like
Interested to know the wider view on what the next priority is.
Assuming the pen enforcement does what we are expecting (enforces min category based on ZP cat). What should the next development be?
- Autocat
- Organiser defined (enforced) pen boundaries
- eFTP to replace ZP w/kg
- Points system
- New ranking system
- Toolkit to use any of the above inputs (options would need to be added over time)
- Something else (explain in comments)
0 voters
2 Likes
I voted for new Ranking system because I assumed that’s where results based promotion would be placed?
1 Like
2 Likes
By points, I assumed the other James meant ranking points - not some kind of league where Zwift decides the divisions and promotions. I wouldn’t like to see that at all.
But I wasn’t disagreeing with you - pens should always be organiser-defined and, yes, automated allocation to apportion each pen with some kind of spread that isn’t based on ability is both unnecessarily complex, confusing, unhelpful to good racing, and leaves racers in the dark until the last minute (I addressed this already a few days ago as a very bad idea).
3 Likes
I don’t think we have data to back this up. Look at any B race result and you can see a clear pattern of higher wkg riders finishing with better positions. If the wkg is a big factor in determining speed in the game parameters - which it is - then race results will also be determined largely by wkg numbers. That’s just how the game works, whether we like it or not.
2 Likes