Introducing ZwiftRacing.app - Zwift Racer Rankings Done Right

Racing
123

Bonus points for podium not quite what I meant.

I used to mention ZP method all the time because I felt I understood it. I now favour this one constantly moving ranking system.

But to use ZP as an example as a general rule, with all riders equally ranked ( never happened) the winner would gain 10% ranking points, middle placed rider no gain at all and last place would lose 10% ranking points. All those in between would gain or lose a proportionate number of points.

I’m not saying 10% is right, it is probably far to high.

I’m just asking the question, which I believe OP asked us to do, whether the current 1 or 2% I’m generally seeing is enough.

124

I would want to incentivise people to keep racing, if you know the bottom half are not scoring positive points it kind of encourages people to drop out or carefully select their races.

For me, you should score points against who you race. For example, If I raced against a stacked field and I was one of the lowest ranked riders I could finish in the bottom third but still score positive points. It means there are still races all the way down through the places.

125

If it’s like chess ELO, it’s not so much about the lower half of a pen’s finishers losing rank points, it’s all about what their ranking score was before the race relative to those who finished ahead.

ELO rank scores, when they accurately reflect the ability of the competitor, gives a probability of who is expected to win a match. If two riders with equal ranks scores of let’s say 3200 race each other, after the race the rank score change will increase for the winner and decrease for the loser by the same amount eg. +5 / -5.

The rank score range in a pen will alter how big post-race changes are. If the lowest ranked rider comes last in their pen, that’s expected, relatively speaking they won’t lose many points and everyone who beat them won’t get much of an increase for beating the model’s weakest rider.

3 Likes
126

For me, I feel it would need a sliding scale of how points are applied - you score more points the further down the rankings you are as you want people to progress quickly.

you will have people with no experience who are simply better than 90% of the rest, how long do they need to take to get where they should be. If they are consistently winning and are the best racer in the race, it minimises the points they can score. Especially if you go the full hog and use rankings / score to set category boundaries

127

I agree with what you have said.

But my question is IF (and a big if) everybody in a 100 rider field started with exactly the same rank somebody has to gain the most and somebody has to lose the most - that is a ranking system at its simplest. ( It gets so much more complicated when riders are not same rank as it becomes, as you say, ranking points then based on where you were expected to come.)

So all I’m keen to know is what is %max gain and therefore loss.

128

How about seeing it as losing less points the better you finish in the bottom half of the race.
Also as said above race hard to beat your expected finishing position and gain points.

I think there have to be negative points otherwise the ranking of everyone just goes up and up and up. I’m sure there are systems like that where they regularly just have a points reduction reset.

129

got bored at work, so looking at the race 2 rankings and comparing them to ZP rankings, my ZP ranking is 303.75 and 3542

got beat by riders ranked with current ZP Cat and race ranking (before the race results)

311.20 - B - 3573
318.54 - A - 3651
223.99 - A - 3957
129.57 - A - 3945
275.28 - B - 3931
170.11 - A - 3943
177.06 - A - 3812
234.05 - A - 3611
272.93 - A - 3769
191.18 - A - 3712
216.78 - B Almost A - 3804
218.03 - B - 3718

whereas I beat riders ranked

245.68 - B - 3602
282.08 - B - 3505
184.64 - B - 3801
295.23 - B - 3551
233.11 - A - 3597
218.37 - B - 3647

podium finishers of that race went up by 50.96 points, 39.36 and 11.72 respectively

now I know I didn’t get anywhere in that race 13th from 19 but I did beat a few riders with higher rankings than mine, I would have expected a slight increase in ranking not a decrease?

130

A very quick look at my race history on the ranking site shows…

https://www.zwiftracing.app/results/3122879 was my biggest gain, 2nd of 96, pre-race score range of ~1600, +76.40

https://www.zwiftracing.app/results/3122875 (the Tiny Race immediately before above) was my biggest loss, 46th of 94, pre-race score range of ~1900, -53.83

The gain race flatters me an awful lot, so many riders still ride dirt route races on road frames and wheels instead of gravel frames and custom gravel wheels, I passed so many riders on “wrong” bikes up the dirt climb.

Greg Alexander is a massive anomaly is the above races, he has dominated the C Tiny Races and his rank score of ~3700+ shows this rather well!

131

Agreed there should be negative points, I just feel that should come from finishing behind people who are ranked lower than you not from finishing position.

You race people, not positions.

If you race positions it always leads itself to be gamed.

132

You shouldn’t score more points just because you are lower but what Tim as looked at and not sure how well implemented at all in the current system is a k factor where if ranking does not have much history on you and you perform well in a race you will move faster.

Over time if you are consistently around a certain rank the movement will become slower.

Lots of good discussion on the ranking discord if you want to provide further feedback ZwiftRacing.app

The one thing with all of this is until zwift force it it will never take hold. Those riders that are ranked way higher than they are on zp have avoided the ranking races from what i’ve seen. They like to sandbag in zp races and similar here if they can’t be on the podium they don’t like to race.

1 Like
133

We spoke in the Discord channel :wink:

2 Likes
135

We have next Ranking Trial Race ready for Wednesday 2nd November ar 19:05 UTC. Please check the description be fore you join.
ZwiftRacing.app Rankings Trial Race 3 - ZRL Recon ZwiftRacing.app Rankings Trial Race 3 - ZRL Recon - The at Home Cycling & Running Virtual Training App

136

Does the “or WKG limit” mean your ZP 20 minute power ? If so then my one-off 3.7 in the first lap of a Bologna RR means I’m going to feel right at home at the back of B.

137

You should be in Pen C…Ranking puts you in C and your PB 95% of 20min is 3.53W/kg.
Enjoy your racing

138

Thanks, I’ll probably be warming up in B climbing glass mountains in your Sunday race :sweat_smile:.

1 Like
139

ZwiftRacing is upgrading!

@xflintx (tagging as these learnings will no doubt be interesting to you too)

Now that we have had a few weeks with the rankings and have been able to analyse some big data sets, we have spotted a few areas for improvement that Tim has implemented this week. These change things considerably.

Change 1 - New Ranking Algorithm

Previously the rankings used an exponential component which heavily weighted finishing position when assessing a rider’s race result. What this meant is that riders that finished highly typically ranked up, and those that finished towards the bottom of the field lost rank. This had some undesirable consequences:

  1. riders would lose rank even if they overperformed versus expectation

  2. riders that raced frequently in weak fields would be promoted far above their actual ability (cat C riders with Grand Champion status)

  3. In races with a large number of participants, ranking changes were inflated

The new algorithm uses a linear calculation that means in almost all cases, you will rank up if you exceed expectation, lose rank if you perform below expectation, and have minimal change to your rank if you meet expectation. This will still promote riders that consistently win, but will further improve the accuracy of the ranking, which is critical if used for pen definition.

…

Change 2 - New Initial Seeding Methodology

One of the most difficult choices with a ranking system is how to deal with new or returning riders. Whilst the algorithm can adapt a rider’s rank quickly in the first few races, it is still important to get that initial placing as good as possible.

The old system used 20m w/kg from 90 day race history. For new riders that may mean only 1 race, so often this would not be a very accurate initial placing. Alongside this, lightweight riders would end up with an overly high initial ranking, and heavyweight riders and overly low ranking.

In the new version, Compound Score is used. This is a calculation of 5m power using both raw W and W/kg to neutralise the weight element. 5m power is also much more reliable in terms of applicability to Zwift. Note this is the only time power is used as a component, and the rankings quickly adapt after the initial placement.

The new rankings are currently populating, and will probably take a day or so to complete. I am planning on arranging another ranking-based race once this is completed and will post here when it is done.

Note that the ranking values and boundaries between ranks are now different, check out the site for info.

ZwiftRacing.app

If you are interested in discussing the nitty-gritty parts of this, the discord channel is the place to be.

15 Likes
140

I admire your optimism that Zwift will take any notice whatsoever of the fine work that you’re doing here on this.

1 Like
141

There is a huge amount of potential with it…

1 Like
142

If someone wins all the time or always places top 10, continuing to do so would mean “meets expectations”?

143

I stopped looking, once I realised that my ranking was dropping for every race I did :rofl:

Having said that, I just took a quick peek and it looks like the rankings may have re-set recently and I’ve “improved” again (won’t be frightening the podium anytime soon)?

No rating or points showing for any race I did in October or November, so basically I don’t have a clue how it operates. If it becomes the Zwift default, I’ll invest more time in schooling myself :blush:

I wonder when Zwift’s new ranking system will make its debut?