Discussion: Points for events - uploaded results

Was looking at my event scores to see any general changes and noticed the last two events didn't have a value:

km pace pts pl
Oct 30, 2011 ROC - Powder Mills Park [2011 Series #6]:Red 7.2 11:37
Oct 2, 2011 ROC - Mendon Ponds Park [2011 Series #4]:Red 6.3 9:29

Looking - I see only 3 of the competitors claimed their events in AP, but entire results were loaded in (and I'm very appreciative of that).

So - it appears the "4 to count" rule is in place as "4 claimed to count" - but the results are all there. Can it be changed to "4 entered results" to count?

If the events aren't claimed, and can't be associated with people who have other results, then they're meaningless number floating in space. Consider the degenerate case where you are the only person who has claimed his splits. What value should be assigned to the event? Makes a difference whether the other times are from people who are good or not, and if they're unclaimed, there's no way to know.

...if they're unclaimed, there's no way to know.

Sure there is.
Each result has a name.
Each name can have multiple results, despite none being claimed on AP.
AP could track those unclaimed results on a per name basis and and use them in the ranking calculation.

But that's hard to do automatically, because you can't rely on unclaimed names being spelled the same way between events. Take my own name: sometimes it will show up as "J-J Cote", sometimes "J.J. Cote", other times as "JJ Cote", and maybe occasionally as "Jean-Joseph Cote", or "Cote, J-J" or whatever. Can of worms.

True. All of those JJ results would be assigned to a different "person". But the system would work most of the time and would be better than just ignoring a majority of the results.

As you guys have said, it gets complicated, but might be possible to do something without letting any worms out. I will have to experiment.

Ah - thanks for the conversation - I haven't had a look under the hood and hadn't considered the issue. Didn't know the ranking was relative across other events.

Let me ask this - what would happen if 1 experienced person was in an event with 3 others who had no other race experience? (What would the effect be of creating "shadow" competitors that just had this one race attached - even if no other name matching could be done?)

The competitor with a ranking would earn their overall average score, independently of how they actually did in that particular race.

(This is not exactly right because of some subtle details about the reported ranking not being exactly the average, but it's basically right.)

And as the others gained experience in the future, it would feed back and adjust the score for the race (and therefore everybody's rankings), if I understand things correctly.

Indeed.

And I know of another David Levine whose results sometimes appear in QOC results lists... How would the rating system differentiate this?

Hmmm - so I have some familiarity with the overall O-ranking-problem, I just didn't know AP had it's own implementation - very interesting (and complicating;-).

To answer David - people with the same name would be averaged together if they were combined...

And I suppose I have the answer already, but if a competitor changed abilities over the course of time that ranked entries were included, how would that complicate things... Average-Joe has a ranking race in May, and Good-Joe has a race in September; the system can't tell that the competitor is better in the later race, and must calculate as if a competitor with the same capabilities was racing in all situations...

Ok - I retract my request;-)

Well, there are any number of people who wildly oscillate between good and terrible. The idea is that if there are a lot of ranked people at a given race, and if most people run a lot of races, that effect gets damped out. Small data sets are always dangerous in this situation.

But performing well and not-so-well is different than actually changing abilities.

If Average-Joe gets beaten soundly early in the season by Good-Fred, but then improves and Good-Joe beats Still-Good-Fred late in the season, the system has no choice but to say that Joe's performance is centered somewhere in between the two events and he's erratic instead of performing to his abilities - that have changed.

Perhaps it is the same, but erratic performance is different than changing performance. It may look the same though, so I'm not pushing anything here;-)

And an injury can take things in the opposite direction.

A way to avoid the hassle of taking care of different spellings of a persons name, several people with the same name and girls (usually) changing last names would be to use a unique identifier (AP userID?) for each person. This of course would require each person to supply this number when entering a meet...
This is the way it´s done in the Swedish ranking system - but that is/was a paid service originally used exclusively for elite runners. The system has changed recently so I´m not fully updated though, but it now includes all runners... This can be done since the Swedish system requires exactly this unique identifier for each person that gets entered through our Eventor system.
Somehow you´ll eventually have to manually attend to the situations above and correct mistakes that appear to get a correct list in the end. People are of course expected to report errors so they can be corrected - this happens quite often, but at least earlier those who didn´t really care about the rankings could appear under several different names.

The process is already automatic for anyone who has an AP profile with an SI number listed. Most regular AP users don't need to claim their splits, because the uploaded results file includes the SI number. The first time someone claims splits, they get the opportunity to save their SI number, and AP recognizes them in the future.

And my particular situation started with non-AP members - we have a good set of regulars that use AP, and a healthy mix of local competitors that do not. The issue isn't getting AP members to claim their races, but that the event has a dozen more that probably will never;-)

...and what if they use another SI (or someone borrows their stick) - that happens from time to time?

SI seems to be a slightly less unique personal ID number.

Neither of those things are currently problems. People can still manually claim splits from other cards, and results on borrowed cards don't get assigned to the wrong person because in addition to saved-card-number, AP also checks that the name on the result matches a name that you have ever manually claimed results for in the past. So, runners do need to claim results manually once per spelling of their name. But it mostly just works.