Discussion: Attackpoint Rankings

Is the system going through an upheaval? Or simply a bug at the moment? I see some people have single digit figures as their ranking, and upon further investigation, pdady's remarkably impressive ROC results look to have shifted the whole attackpoint population?

Human (manual input) error, nobody can run course in 1:20 (1 min. 20 sec.) while other same course took more than 1hr.

I noticed the other day that rankings seem to have issues at the moment.

I sent him an email in case he doesn't see the notes in his log.

Has the system for calculating ranking points changed recently? My last two runs I finished 1st today (with a good run) and 2nd a couple of weeks ago (with a terrible one, only a few seconds behind the winner) but points were both in the 40s whereas today's sort of run would normally get into the 60s.

There haven't been any changes to the system.

Maybe I needed Craig there to boost me up the points list.

Surely our minor events don't result in big ranking points for tRicky?

No big ranking points but I'm used to scoring in the 60s for our local events with a good run, not that 46 pittance - typically a national league event against the big guns I'd be in the 70s if I didn't muck up too badly and I don't expect any higher than that.

I just recall the ranking system changed many years ago because our local events used to score up into the 90s so thought maybe it had changed again.

More likely that the system hasn't had any/many events fed into it lately, for obvious reasons?

It has definitely been a bit off for a while. Everyone's points generally seem to be a bit lower, outside the top 2, the rest of the top rankings are much lower than I have seen in the past. And inside the top 2, BigE & Rhombus' rankings are unexplainably high - Rhombus doesn't have any race with as high ranking points as his actual ranking.
I (and many others I suspect), currently don't have any ranking points, and I'm not sure why.
Something has shifted the points downward in general (and upwards in the Canadians' cases), and I would suspect the downward shift has affected your results tRicky. Could be related to the fact that pdady's previously mentioned results are still remarkably highly ranked (despite the human entry correction).

Interestingly, the top three, BigE, Rhombus, Flakey are all from the same club in Ottawa. I would guess they've somehow pushed each other to the top and inflated their scores.

Attackpoint rankings are more robust when the 'connections' between different groups of people are more frequent. In the situation where there are 200 people, of which 100 in group 1 and 100 in group 2) and 1001 races:
- in 500 races, only group 1 people race among themselves
- in 500 races, only group 2 people race among themselves
- in 1 race, 1 person from group 1 races against several people from group 2,

the one result of the group 1 person in the shared race will be crucial in ranking group 1 against group 2. The Attackpoint ranking algorithm assumes that the shared race exactly measures how good group 1 is against group 2. So, if the group 1 runner has a bad day that day, group 1 will be imputed to be terrible. If she has a great day that day, group 1 will be ranked very highly.

This effect of one bad day crucially affecting lots of rankings occurs when connections across groups are weak. As happens in a pandemic, when fewer races are within the 1-year window that AP uses in the calculation.

Therefore, likely what happened is that BigE or Rhombus or someone they race against in Ottawa has a very good result or two against someone not normally in their group, and this causes the algorithm to deduce that Ottawa is the center of the orienteering world. Because the top three runners' average ranking is normalized to 100, this lowers everyone else's ranking.

Actually just a quick look at the past results for Orienteering Ottawa will show you that at least during 2019 Ottawa was pretty well the centre/ center of the orienteering world or at least that part of the world in North America with Robbie Graham and several guys who seem to be too busy to compete for national team spots fighting it out week after week. Then throw in a great mix of XC skiers and adventure racers on fine venues and courses set by well-trained course setters and you have what are likely the most competitive local races on the continent.

I still have a 4.6 point race from pdady entering mm:ss instead of hh:mm for a couple of races. I see that he does not have an overall score, but those races still have him listed with 280+ points, even though the time has been corrected. For example, here:
https://www.attackpoint.org/racesplits.jsp/race_34...

@bishop22: that race is not currently listed in the rankings, suggesting that it has been flagged in the background as a training event or otherwise as bad data and is therefore being skipped in the calculations.

The Attackpoint ranking scores computed on race pages are not recalculated as frequently as the points on the main ranking page. The scores you see were presumably the points last calculated before that race was flagged as bad data. (ken can confirm.)

I did follow that up with Ken a while back and he confirmed it was removed from the overall rankings (edited as a 'training' race).

Regarding our local races, I just found it odd that a similar local race with similar competitors earlier in the year scored me 69 points whereas the race on the weekend (both of which I won) only got me 46 points. The only difference being that Craig was missing from the weekend just gone.

I think sometime in the past couple of weeks last year's Canadian Orienteering Championships were dropped from the calculations of the rankings, which meant that all of sudden there were very few comparison points between eastern and western Canada, and also a big decrease in comparison with the US, and by extension, down under.

In fact, if my quick look through the ranked races is correct, the splits used to calculate the top three, BigE, Rhombus, Flakey, (all from Ottawa and using the same 2 or 3 races for their splits calculation), only have one connection point with the rest of the AP community, that being sberg (also from Ottawa, but who races typically on a different course). sberg raced against LKohn and RWorner at a singular race in Ottawa, and beat them significantly, which tells the algorithm that the top three are significantly better than sberg, who is much better than LKohn and RWorner. LKohn and RWorner have competed in several races in the US, and from there the rankings spread out - with BigE, Rhombus, Flakey at the top of the rankings.

Time will tell. The top three have no ranked races past Oct 2019 so perhaps they'll drop off the list in a couple of months' time.

Would it be appropriate for the Attack Point Rankings to suspend the event drop-off for a bit?
For the Florida JROTC team rankings we normally drop points earned after 12 months. We have not dropped points since last January.

I vote to keep the current rankings

I vote to let me keep my ranking points from the last time I raced in an Ottawa OC race against the Ottawa guys (2018) forever. ;-)

https://www.attackpoint.org/racesplits.jsp/race_28...

Why Hammer? Is that the last time you got a better result than Adventure Girl?

I like that you guys are voting when attackpoint is clearly a dictatorship.

The IOF reacted on this thread:
https://orienteering.sport/wre-period-prolonged-du...

It's the only way to raise an obedient child, yurets. Thought you'd know that.

I don't really understand how the ranking system works. Yes for individual races I kind of get it but my current ranking score is 49.3, yet all but two of my races in the past year have scored significantly higher than that (10 races scoring from 64.9-77.0). It's only the most recent two that have scored 42.8 and 46.0. The people on either side of me in the ranking list have most of their scores around about their overall ranking score so it seems a bit weird in my case.

The tricky thing is that scores for your races will vary depending on the ranking period/window, meaning that scores can change for a year after the race, until they are no longer included in the ranking period.

At that point the race scores are influenced only by the races that happened in the year afterward, and none that happened before. This is the opposite of what you really want to remember about any particular race. So we freeze the race scores displayed on the splits and your profile a few weeks after the race.

But the race points used to calculate your current ranking do continue to float after that based on all the mysterious behaviors inherent/emergent in the algorithm.

So tRicky, you need to look at your race points here, they have changed.

Well poo.

The tricky thing...

What you did there - I see it.

The Nakkertok and Bilberry Creek races were fire! Parc Moussette is a little suspicious though, a long sprint race where I didn't even warm up... gets a casual 100 points X'D

BTW The club has some sharp new Gatineau maps in the works. When events start I personally invite any Americans to come up and we can race! Could be some big points in it for you ;-)

It seems like it'll be pretty hard for me to make a trip up there for a while, unless I can find a quiet spot on the border and get good topo maps on both sides...