Introducing ZwiftRacing.app - Zwift Racer Rankings Done Right

DNFs…

Thinking back to Microsoft Gaming Zone, that used an ELO system, players who failed to complete games were given “incs” (incompletes). The number of incs in a player’s MGZ history was there for all to see. Every time a player accumilated five new incs, the ranking system treated those five games as losses and adjusted rank points accordingly for all involved in those games.

The odd internet/other issue gets some leaway, serial DNFers get some punishment.

10 Likes

Great idea, I will mention it in the discord

Is it a punishment if their goal is to drop rank so they can again be the hammer? Just go in and start/quit 5 matches, drop a rank, and then smash a race.

The more I think about this, the more ‘shifting’ pen definitions need to become the norm rather than static A/B/C/D/E definitions that can be gamed over time, that way it’s not super productive to try to shoot for a specific ranking to keep you top of a specific division.

3 Likes

I agree. A good portion of riders are just looking for podiums and don’t care about rankings other than as category boundaries that can get in the way of podiums. They don’t even necessarily care about podiums on ZwiftPower if the Companion app says something better.

1 Like

I don’t know how much emphasis I can give to this point, but flexibility in pen boundaries for race organisers, regardless of the metric used to sort them, is THE KEY REQUIREMENT. Matchmaking would be even better.

3 Likes

@gloscherrybomb @TimHanson This ranking system looks awesome :+1:

I’ve not used Zwift for a while (moved to RGT early 2021), but if Zwift incorporated this ranking system (or introduced something similar), then it could definitely tempt me back to Zwift.

2 Likes

That’s cold hard cash right there :point_up_2:

1 Like

We agree on that, James. I have been saying this for a long time too. In fact, series like Mad Monday have been using custom pen boundaries for a long time, and we are planning to take it to the next level, soon. All scripts have been finished already (but awaiting Zwift HQ clearance for one script).

1 Like

What’s the next level, pen enforcement? Maybe TFC could do a series sorted by rank rather that power metrics? A big race organiser buying in would really help push the case. There is an API now to pull ranks using ZID which makes it a lot easier to determine who goes where.

2 Likes
6 Likes

Quick question: Does it rank based on primes if those are processed in Zwiftpower? Just curious as results seem to imply that when looking at a specific race. If so, that’s great! However might be like current ZP rankings which are just based on finishing even though positions can be different based on primes.

Problem is only other competitions, such as ZRL, which are not properly accounted for on Zwiftpower. Had most points in the first race but only rolled over finish middle of the pack which hurt my rankings here quite a bit. Can’t blame these rankings as it’s the bad setup of Zwift but not perfect either (again, finger to be pointed to Zwift).

Rankings seem to be much more sensible than before (no B riders on top of the world anymore, nor C’s with very high ranks).

Points races are supported if they are set up properly in Zwift Power. That’s why you see some ranked properly and others not. ZRL support is going to be investigated for a future release, it’s tricky because they don’t set it up properly in ZP.

Does or can the ranking system deal with different types of races? A rider would be rated differently on flat vs hilly vs real climb, also taking into account the length of the race and the nature of the finish (flat finish vs uphill)? Would a race route need to be rated itself?

2 Likes

No, but this is asked quite often, and people have different views on it.

For me it’s important to understand the purpose, the whole philosophy, behind the rankings and why this approach was taken.

It gives people an accurate rank so that when pens are formed based on rank, they are fair and fun. It gives riders something to fight for in every race regardless of whether their expected performance lies. It rewards improvement. It is results and performance based. It is updated in an accurately way regardless of how the pens are sorted (it is not reliant on pens being sorted by ranking itself).

It is not a way to rank rider by capability. It is not a way to sort all riders from best to worst. If that is what you want, you just take a ZP-like approach where you only consider the best performances. However, this approach is not compatible with the benefits listed above. That’s a concept that took my a while to get my head around, so I can explain it further if needed.

Once that is understood, you see that the ranking system really has to be global - all results, all positions in races, need to be taken in to account. If a sprinter only enters sprint races their ranking will natural be higher than if they entered a range of formats. However it is accurate, for the races they take part in. If they enter an AdZ race and do badly, their ranking will adjust downwards. So it is always self-correcting - and when used to sort pens, even if there is a relatively wide range of capabilities, it is fair…

So no, there is no separate ranking for race type (once you think about it, you start to realise that would be extremely difficult to do anyway. Sometimes a hilly race is won by a sprinter in a big pack, sometimes by a climber, sometimes by a puncheur).

That said, one of the benefits of ranking using this type of approach is that the way pens are sorted is really irrelevant. It could be completely random. It could be power based. It could be age group based. It could be ranking based.

So if you are a race organiser setting up a hilly race, you have plenty of options depending on the type of experience you are trying to create. Examples:

  • Split by ranking
  • Split by 20m w/kg
  • Split by 5m Compound Score
  • Split by rider phenotype
    etc etc

Apologies for the long answer to a short question!

1 Like

Thanks James, and I appreciate the “long” answer … it helps me (everyone) get a better understanding of the system.

It certainly looks like an interesting and valuable development.

Can you say what would happen in the simplest possible idealised case where all riders have a true fixed performance rank, and the pen boundaries are not altered from one race to the next?

Ie say rider 1 always beats 2-100, rider 2 beats 3-100 etc. assume a weekly race with 2 pens with the boundary set between 50 and 51. How often does 50 get to lose before they get demoted and start to win the lower race? Does 49 always finish next to last in the top race?

In that scenario, as it is set up today, you would see very little movement throughout other than the podium positions which would immediately move up (due to the podium bonuses). And of course, if they move up based on your example, the bottom placed riders would also move down.

The general take is that if riders are doing exactly what you would expect them to do race after race, there is no need for the ranks to be altered. The podium bonuses are there as a bit of a sticky plaster to deal with the repetitive nature of zwift races using the current pen boundaries to ensure riders ‘stuck’ at the top of a category move up.

So if you are expected to place just outside the podium, and place just outside of the podium, you will not be bumped up. Is that correct?

In the purely hypothetical scenario described above, correct (bumped up doesn’t really mean anything, unless you’ve contrived some format like the one James has used as an example)

1 Like

So does this mean you have points inflation due to bonuses? I’d have thought it best to try to conserve ranking points across the field, perhaps by trimming the ranking of everyone in a given race enough to compensate for the bonuses. That would tend to push a few weak riders down over time, such that they’d drop into a lower race.

It sounds otherwise like there could still be a load of people stuck at the bottom of the pen with little prospect of moving down into a more suitable event any time soon. (I agree that varying pen boundaries will help with this, but they have to vary enough, and this has to be effective enough to handle riders coming into and out of the system with unrealistic ranks.)

I really don’t think my example is that contrived, it’s surely the most obvious starting point for any system design. What other simple and easily understood scenario would you prefer? The point is that it’s sensible to make sure that any system copes adequately with the simplest possible challenges before expecting it to work well in reality.