Unsure about crews going down a division to position 2

mcshane-fire commented 1 year ago

In Lents 1986 mens results, on day 3 at the bottom of division 3, Sidney Sussex II goes from the last place in division 3 to the second place in division 4. Normally this can't happen, as only a single sandwich crew would row at the bottom of division 3, so they could only be bumped by a single crew. Here's the chart:

A similar thing happens a few times in torpids:

1984 men, Christ Church II on day 3
1993 men, St Annes on day 3
2019 women, St John's II on day 2

Do you know the reasons behind this? Was this the result of a penalty? Are the ones in Torpids the result of allowing the top two crews from one division to race at the bottom of the next division up?

mcshane-fire commented 1 year ago

Or is the Lents 1986 example because of a reporting error, and actually they were bumped down by Christs III on day 3 and bumped by LMBC V on day 4?

mcshane-fire commented 12 months ago

I found these by seeing if my results format can cope with Torpids. Apart from the 60s and 70s where the division sizes changed each day, these three occasions are the only problems. I did have to fix one bug in my results parser script, which didn't cope with a crew bumping up to sandwich position and then having an explicit results (e1) in the next division up that day.

As an improvement I added support for a shorthand for the '\<n> crews go up one, and one crew goes down \<n>' which is common in torpids, I used 'w\<n>' as the results code, which replaces 'e1e1e1e...e\<-n>'. I've got code that converts from ad to tg format, and code that parses tg results and creates charts, but I think you re-implemented this all so I'm not sure me sharing my changes is useful. If I add support for division sizes changing each day, and we figure out what's going on with the four situations listed in this issue I can push all the torpids results in tg format if that's useful.

johnwalley commented 12 months ago

I'm not sure about the Lents 1986 mens results. I'll have to go check the original sources.

For the Torpids results I tend to assume there was a penalty applied but in general I don't know. The official results match up for my 2019 women's results:

as do the official start orders for days in question

My hypothesis is that St John's II picked up a penalty bump on the second day which moved them down the start order for day 3 but I can't find any evidence.

I'd be interested in seeing support for changing division sizes. It's something I've thought about adding but not got around to yet. And given the ad format doesn't really support changing division sizes (it keeps all the crews in the right positions but doesn't convey any information about division sizes to any visualisation) I could use the tg format again as the canonical format.

mcshane-fire commented 12 months ago

I've created #13 for handling the Torpids 60s and 70s rules which changed division sizes each day, to keep this issue about crews going down a division to position 2.

2019 is recent enough so we should be able to find someone from that crew and ask them!

mcshane-fire commented 11 months ago

I sent Anu an email, he agreed that these crews likely received a penalty bump, so putting them down an additional place.

My results processing does process one division at a time, so rejects using exact moves (e codes) that move a crew down a division below head of that division (since that would normally indicate an error in text results). My proposal is to add an additional results code 'p' that means 'give the crew that gets the next exact result code an additional penalty bump down'.

I checked this on the torpids 2019 women results, here's the bumps chart around St John's II that do this on the second day.

The results for that division are: e1pe-1ure2re1e-3urr The top few crews in the previous division must have all rowed over. The results are read to mean:

e1: Sandwich crew (Wadham II) goes up one place
p: Next crew will get an additional penalty bump
e-1: Crew 12 (St John's II) goes down one place to Sandwich (or head of next division down)
Apply penalty bump: They move down an additional place from head of the division to second place, with the second place crew (Somerville II) going up one.
u: Crew 11 (Corpus Christi) bumps Crew 10 (Oriel II)
etc...

torpids2019women

torpids2019_women.txt

The syntax also allows for multiple penalty bumps (so two penalty bumps applied could have been: e1ppe-1), with every crew they go down past going up an additional place. The penalty handling applies to the next crew that has an 'e' result code. If that crew that received a penalty bump rowed over then you can use e0 as the result code.

For penalty bumps that don't cross division boundaries we can't normally tell from a bumps chart whether the bump happened on the water or as a penalty, so they would be coded without needing the 'p' code. The only place this is needed is when the results are rejected since the crew drops down a division lower than the head of that division.

johnwalley commented 11 months ago

I think I follow. The wrinkle in processing is now we have to potentially go back a division to put crews in the correct places.

All the additions to the format look good to me.

mcshane-fire commented 11 months ago

I've checked in my results files that include the p code for this special cases https://github.com/mcshane-fire/bumps. The readme.txt file documents the file format. For now I've duplicated the results files, we can clean that up later by deleting one set of duplicates or just each maintain a set of results files (which are likely to end up identical).

johnwalley / bumps-results

Unsure about crews going down a division to position 2 #8