Zwift Chat

[<Tommy>]

Just logged in and there's a banner at the top that reads 'SBP Ranking in progress. ACE Scores shown are not final.'. Everyone on the list has just 'Calculating...' to their name.

But it does make sense for me and Paul to be somewhat close. Last season of ZRL, excluding TTs, Paul got 3rd, 8th, 13th & 22nd. I got 4th, 9th, 14th & 26th. Pretty sure I outscored everyone on FTS/FAL points too. Paul's best 20 min W/kg were 3.94, mine 3.89. Paul did 30 events in the last three months, I did 38.

Dunno…

Your results last season took a big upswing. But the results have been taken over the last 4 months. If you compared yours and Bridgy’s results and power over that stretch it would look somewhat different wouldn't it.

Alex is also higher ranked than a lot of the B cat cyclists as we’ve already mentioned. So maybe these things combined are little hints as to how the ranking is calculated. I wonder how the scoring value differs between the categories. And what value they put on each aspect. Maybe it’ll settle down over time. And become clearer too!

I am genuinely interested. It does all seem overly complicated though doesn’t it?

 

[Peter Salt]

Dunno…

Your results last season took a big upswing. But the results have been taken over the last 4 months. If you compared yours and Bridgy’s results and power over that stretch it would look somewhat different wouldn't it.

Alex is also higher ranked than a lot of the B cat cyclists as we’ve already mentioned. So maybe these things combined are little hints as to how the ranking is calculated. I wonder how the scoring value differs between the categories. And what value they put on each aspect. Maybe it’ll settle down over time. And become clearer too!

I am genuinely interested. It does all seem overly complicated though doesn’t it?

Might be simple or complicated. Without seeing it, we don't really know.

But what you pointed out is useful - it seems like it takes into account results/numbers from a long time period, but if they recently improved then the ranking will reflect that. Just like ZP calculates your category based on 90 days of results, but if your 3 best efforts were in the last 2 weeks, it makes the previous ones insignificant.

Probably a good thing.

 

[<Tommy>]

Might be simple or difficult. Without seeing it, we don't really know.

But what you pointed out is useful - it seems like it takes into account results/numbers from a long time period, but if they recently improved then the ranking will reflect that. Just like ZP calculates your category based on 90 days of results, but if your 3 best efforts were in the last 2 weeks that makes the previous ones insignificant.

Probably a good thing.

Yes but the two things contradict each other don’t they. You can’t have a points system and a system based on best 3 results can you. That’s one thing that seems odd about it.

Obv the current cat system is power based. So it’s easy to do the 3 best results thing. But assuming a points system allocated points on a per race basis, and the points are culminated to contribute to your rating. Then it doesn’t make sense does it? That’s what I was questioning in an earlier post. How have I for example accumulated 750 points. See what I mean?

Personally it looks to me like Alex should be racing in B. That’s his choice… But the points he has amassed compared to lots B cat cyclists would also maybe imply a few things about the way they are spreading / valuing different aspects. Eg points vs power. And also the point spread differences between the categories.

Anyways… I’m definitely spending too much time thinking about it!

 

[Legs]

Stephen King isn’t celebrated enough (In my opinion)

Just started Duma Key, which is the forty-first full-length Stephen King novel I’ve read (that’s not counting long novellas like Rita Hayworth and the Shawshank Redemption, 1922 or The Mist, which appear in collections of short stories, of which I’ve read seven). I guess I’m what he would call a Constant Reader! My favourites are Dolores Claiborne, The Green Mile and the title novella from Hearts in Atlantis. 😃

 

[steverob]

Obv the current cat system is power based. So it’s easy to do the 3 best results thing. But assuming a points system allocated points on a per race basis, and the points are culminated to contribute to your rating. Then it doesn’t make sense does it? That’s what I was questioning in an earlier post. How have I for example accumulated 750 points. See what I mean?

The trouble with points only systems is that they are very easy to game by having a couple of poor results in between your good ones. Take the ZwiftRacing.app system - that's supposed to be based purely on your results (as far as I am aware), but I had two races where I had results the system clearly regarded being as poor for me and my rank massively dropped after both of them.

On neither did I deliberately attempt to get a bad result - one I went round with the lead group but didn't contest the sprint cause I was going for a GC time, so dropping 30 places was not a concern for me as it only cost me one second; while the other was one where I'd already done an event an hour earlier so I was already tired (though I did push as hard as I was able to), but I needed to compete in both as they were part of larger events and it was the last day they were available. It then took me a good few weeks of decent results to get my ranking back to where it previously had been - it crept back up a few points at a time despite having lost almost 40 points in those two races alone.

Now if I had TRIED to get a bad result then logically my rankings could have dropped even more I guess? So I think that having power numbers somewhere in the formula mix as well might not be a bad thing if it means you're not able to drop below a minimum threshold, otherwise an A rider could just finish dead last in ten races in a row and end up with a ranking that's alongside D riders.

 

[<Tommy>]

The trouble with points only systems is that they are very easy to game by having a couple of poor results in between your good ones. Take the ZwiftRacing.app system - that's supposed to be based purely on your results (as far as I am aware), but I had two races where I had results the system clearly regarded being as poor for me and my rank massively dropped after both of them.

On neither did I deliberately attempt to get a bad result - one I went round with the lead group but didn't contest the sprint cause I was going for a GC time, so dropping 30 places was not a concern for me as it only cost me one second; while the other was one where I'd already done an event an hour earlier so I was already tired (though I did push as hard as I was able to), but I needed to compete in both as they were part of larger events and it was the last day they were available. It then took me a good few weeks of decent results to get my ranking back to where it previously had been - it crept back up a few points at a time despite having lost almost 40 points in those two races alone.

Now if I had TRIED to get a bad result then logically my rankings could have dropped even more I guess? So I think that having power numbers somewhere in the formula mix as well might not be a bad thing if it means you're not able to drop below a minimum threshold, otherwise an A rider could just finish dead last in ten races in a row and end up with a ranking that's alongside D riders.

I’m not sure to be honest.

I don’t think the examples you use are an argument for power and points in a ranking system. I think they’re an argument for using your best efforts over a set time, rather than taking each result on their own merit. Because in those races your power would have also been down as well as your position by the sounds of it.

In a points only system yes, if you get bad results you drop back in the ranking. So in your example, say getting 10 bad results intetnionally would drop you back into D. But as soon as you did try in a race you’d fly back up the ranking again. I’m not sure people would go say, two weeks of finishing last intentionally (10 races = 2 weeks maybe?), just to win one race. And even if they did do that. That is still a lot more effort than it takes to sand bag at present.

For me whatever system they use clearly needs tweaking. I would like a system where it’s harder to go backwards in the rankings than it is to go forward. One way of doing this is giving a bigger spread of points the further up the ranking you go. In the current ZP system the spread is too narrow between the cats. Maybe use power as a sort of hand break. But I think the system should be predominantly points based.

 

[alex_cycles]

The trouble with points only systems is that they are very easy to game by having a couple of poor results in between your good ones. Take the ZwiftRacing.app system - that's supposed to be based purely on your results (as far as I am aware), but I had two races where I had results the system clearly regarded being as poor for me and my rank massively dropped after both of them.

On neither did I deliberately attempt to get a bad result - one I went round with the lead group but didn't contest the sprint cause I was going for a GC time, so dropping 30 places was not a concern for me as it only cost me one second; while the other was one where I'd already done an event an hour earlier so I was already tired (though I did push as hard as I was able to), but I needed to compete in both as they were part of larger events and it was the last day they were available. It then took me a good few weeks of decent results to get my ranking back to where it previously had been - it crept back up a few points at a time despite having lost almost 40 points in those two races alone.

Now if I had TRIED to get a bad result then logically my rankings could have dropped even more I guess? So I think that having power numbers somewhere in the formula mix as well might not be a bad thing if it means you're not able to drop below a minimum threshold, otherwise an A rider could just finish dead last in ten races in a row and end up with a ranking that's alongside D riders.

Power numbers ARE in the ZwiftRacing.app rankings. Your initial score is your compound score, which is your 90-day best 5 minute raw Watts x 5 min Watts/kg.
Even that on its own is pretty indicative of where you are at as a racer. As far as I'm concerned it's the best system out there (of a fairly bad bunch). If you look at the team rankings, ours are in a fairly sensible order compared to ZP etc.

Also, I was told by them that changes of 10 are regarded as "nothing at all" so a shift of 25-30 "points" is not a large movement, although it hurts when it happens for the reasons you stated. Points change when you do better or worse than expected (as you know). But also, they do a "season reset" ~quarterly (most recently 1st Jan) whereby if your score is more than 200 points away from your compound score, it's adjusted back to within 200 of your CS (in either direction).

Either way, it's a ton better than ZP rankings which are just a silly (but fun) game IMHO.
Our CCC team rankings on the new WTRL ACEscore system, when they were first published, matched the ZP rankings very closely, which shouldn't be an enormous surprise, really since the ACE covers the last 120 days of racing and the ZP covers 90.

But, like Tommy, I've spent far too much time thinking about this. Looking forward to Tuesday though.

 

[alex_cycles]

For me whatever system they use clearly needs tweaking. I would like a system where it’s harder to go backwards in the rankings than it is to go forward. One way of doing this is giving a bigger spread of points the further up the ranking you go. In the current ZP system the spread is too narrow between the cats. Maybe use power as a sort of hand break. But I think the system should be predominantly points based.

Totally agree with that. To me it's absurd that a rider in a lower category can achieve a ZP ranking much higher than riders in higher categories. There should be a category-based component to it. e.g. better points for an A race than a B race.

 

[Peter Salt]

At the end of the day, it's also vital to take into account what this score will be used for - in this case it's the 'segment battle points' system, which puts riders up against others with similar abilities on segments. You want that to be dynamic - i.e. if someone does exceptionally well in an event, the algorithm should pick that up right away and 'promote' them quickly.

Looking at the initial numbers they assigned, I would say that fitness plays a role but not a big role. Ketty, for example, is fitter than me and Paul but only had about 2/3 of our ranking points. Likely on account of doing 4x fewer events in the last 4 months, no top 15 finishes in ZRL and not scoring too many segment points in those races. On the other side of that spectrum we had Michael - much lower W/kg numbers but an absolute beast on segments and frequent racer, within a spitting distance from me and Paul.

So my working hypothesis is that fitness has a very low weighting in all of this. Volume of racing and direct performance against others on segments seems much more important. Makes sense if they want to use it for additional segment scoring.

The only thing I really dislike about it, is that when your best rider can't make the event, every rider below him will end up in a mismatched group - the person who would be in 2nd group will be in 1st, rider in 3rd will be in 2nd, etc. So if your 6th rider cant make it - it's a small loss, but if your top rider can't make it - everyone gets shafted.

 

[alex_cycles]

At the end of the day, it's also vital to take into account what this score will be used for - in this case it's the 'segment battle points' system, which puts riders up against others with similar abilities on segments. You want that to be dynamic - i.e. if someone does exceptionally well in an event, the algorithm should pick that up right away and 'promote' them quickly.

Looking at the initial numbers they assigned, I would say that fitness plays a role but not a big role. Ketty, for example, is fitter than me and Paul but only had about 2/3 of our ranking points. Likely on account of doing 4x fewer events in the last 4 months, no top 15 finishes in ZRL and not scoring too many segment points in those races. On the other side of that spectrum we had Michael - much lower W/kg numbers but an absolute beast on segments and frequent racer, within a spitting distance from me and Paul.

So my working hypothesis is that fitness has a very low weighting in all of this. Volume of racing and direct performance against others on segments seems much more important. Makes sense if they want to use it for additional segment scoring.

The only thing I really dislike about it, is that when your best rider can't make the event, every rider below him will end up in a mismatched group - the person who would be in 2nd group will be in 1st, rider in 3rd will be in 2nd, etc. So if your 6th rider cant make it - it's a small loss, but if your top rider can't make it - everyone gets shafted.

I've only just found this, but if you click the information symbol on the WTRL team portal you get this. So it looks like it is mixed power and results and dynamic. In theory, should be good. We'll see.

ACEScore
This score is calculated by Autocat against all other riders in Zwift Racing League using all races (not Group Rides) in the last 120 days and forms a score out of 1,000 (though can marginally exceed this value). The score is determined using weighted factors from previous ZRL Points Races, wins/losses v other racers in both ZRL and non-ZRL races (ELO), finishing position and a riders current abillity/power curve.

 

[Peter Salt]

I've only just found this, but if you click the information symbol on the WTRL team portal you get this. So it looks like it is mixed power and results and dynamic. In theory, should be good. We'll see.

The really vital piece of information - i.e. what are the ballpark weights of each component, is obviously missing

 

[alex_cycles]

The really vital piece of information - i.e. what are the ballpark weights of each component, is obviously missing

Classified data. Top secret valuable IP.

 

[CXRAndy]

I'm not overly concerned by the technicalities, a bit like quantum physics.

It just is 🤪

 

[alex_cycles]

I'm not overly concerned by the technicalities, a bit like quantum physics.

It just is 🤪

Quite. More pertinent is if the live group tracking will work or not. Would be useful to be able to decide on the fly how hard to hit each segment (as that can/does affect finish position). I'm guessing 50:50 flakey.

 

[berty bassett]

This segment racing just seems alien to me - am I right in thinking that if I blitz round a course at full tilt I will be beaten on points by someone just trying on segments ?