How's it work mathematically ?

Discussions on race preparation, race management, sound effects, and other race related software. This includes freeware, shareware and commercially available software.
Post Reply
Rob Shafer
Apprentice
Apprentice
Posts: 11
Joined: Tue May 12, 2015 5:46 am
Location: Phoenixville,PA

How's it work mathematically ?

Post by Rob Shafer » Tue May 12, 2015 3:12 pm

Hello all a quick question I hope someone can answer. I use Raceview to run our wood turners club battle top event,. (Concaved surface Plexiglass walls, four gates the tops bang into each other last man standing wins etc) my question is the club is getting bigger to speed up the event were thinking of building a second arena, so total of eight "lanes" if the software picks the lanes and we award ties after each heat which it will do . Practically and mathematically will it work? My first question is does it pick lanes randomly after it has data to work with? Meaning we run every body once. The next heat point wise who does it match up and in what lane order? I'm worried if there is some logic to it that some people may not get a chance to face everybody on each arena equally. :scratching: also when we're down to one or two guys on the second arena how should their. Finishes be placed 1st 2nd or 3rd 4th? Weird stuff and a programmer input needed for sure but thought I would throw it out there thanks in advance rob



User avatar
Stan Pope
Pine Head Legend
Pine Head Legend
Posts: 6888
Joined: Sat Jul 05, 2003 7:01 pm
Location: Morton, Illinois
Contact:

Re: How's it work mathematically ?

Post by Stan Pope » Tue May 12, 2015 4:40 pm

Rob,

It depends on the scheduling option that you (or the software) select. If you choose the "PPN" option, then, once more, it depends. GPRM is one software product that includes PPN as a scheduling option. For web page generation of PN, CPN, PPN charts see http://www.stanpope.net/ppngen.html.

In its best forms (PN or CPN charts) everyone races everyone else the same number of times AND everyone runs in each lane the same number of times. PN and CPN charts only exist for a limited number of combinations of racers, lanes and rounds. For the other cases, PPN charts are used in which everyone races in each lane the same number of times and opponents are balanced as well as is possible: maximum difference is 1. There is one more case in which even PPN charts don't exist. For these, the software does as well as it can with respect to opponent balance while retaining the requirement that everyone races in each lane the same number of times.

If you are scoring by points instead of times (i.e. opposition balance is essential) and must run a PPN schedule, then I recommend two phases. The first, preliminary, phase should select about twice as many finalists as you have trophies to award. Aim for a finalist count which matches a PN or CPN schedule.

Questions?


Stan
"If it's not for the boys, it's for the birds!"

Rob Shafer
Apprentice
Apprentice
Posts: 11
Joined: Tue May 12, 2015 5:46 am
Location: Phoenixville,PA

Re: How's it work mathematically ?

Post by Rob Shafer » Tue May 12, 2015 6:27 pm

PPN? As old as it is was hoping some one knew Raceview,, I has a 9 point threshold and you're out if that helps



User avatar
Stan Pope
Pine Head Legend
Pine Head Legend
Posts: 6888
Joined: Sat Jul 05, 2003 7:01 pm
Location: Morton, Illinois
Contact:

Re: How's it work mathematically ?

Post by Stan Pope » Tue May 12, 2015 7:13 pm

Rob Shafer wrote:PPN? As old as it is was hoping some one knew Raceview,, I has a 9 point threshold and you're out if that helps
Oops! I missed your mention of Raceview. Very sorry.

I searched Google for references to see if doc was online ... if available, it is way down the list of search results. There is probably nothing directly from the source since raceview.com was abandoned! From comments, it is likely that support for raceview went away several years ago.

I then searched Derby Talk of "raceview" and found several folks who currently use, or have recently used, raceview. They would be good ones to talk to ... PM or email to them might bring 'em out of hibernation.

If you have a couple of printed samples of raceview schedules, I might be able to sus out the internal algorithm for you. So consider that if you don't get results with PM's to raceview users.

Also, you might tell me about your criteria for determining winners. From your brief comments, it sounded obscure. :)


Stan
"If it's not for the boys, it's for the birds!"

Rob Shafer
Apprentice
Apprentice
Posts: 11
Joined: Tue May 12, 2015 5:46 am
Location: Phoenixville,PA

Re: How's it work mathematically ?

Post by Rob Shafer » Tue May 12, 2015 10:02 pm

Ok thanks, so we run it like a derby scored by finish position. I think 1 point for first 2 for second etc. Might be 0 for first 1 for second etc. You are out when you get over 9 points. This is a 36" round dished arena 4 openings (Gates in the walls) the tops are hand made on wood lathes to exacting weights and measures pretty much unlike derby, unique designs are discouraged we want as long a battle as possible so they weigh up to the same weight have to be within a 1/4" of dia of each other and pretty much have to rub each other in the same spot along the edge. They have a string wrapped around the shaft, "123 Go" the string is pulled and they're off banging and slamming into each other as they fall they are gently removed so as to not interfere with those still spinning. SO logistics here, we can have ties. If we let the software assign lanes, you think it's designed to randomly assign players against each other, without considering their score when doing so? but possibly it balances "lanes" out.
and then what should we do when we're down to only 1 or 2 guys on the second table,? I think they should finish 1 and 2 correct? Seems only fair but I'm not sure.

If anybody out there knows the ins and outs of Raceview please weigh in. Thank you all, my boys are grown but I built the track and ran derbies for all of their years and after while they were in Boy Scouts. It's a great activity



User avatar
Stan Pope
Pine Head Legend
Pine Head Legend
Posts: 6888
Joined: Sat Jul 05, 2003 7:01 pm
Location: Morton, Illinois
Contact:

Re: How's it work mathematically ?

Post by Stan Pope » Wed May 13, 2015 1:34 am

Rob Shafer wrote:...If we let the software assign lanes, you think it's designed to randomly assign players against each other, without considering their score when doing so? but possibly it balances "lanes" out.
and then what should we do when we're down to only 1 or 2 guys on the second table,? I think they should finish 1 and 2 correct? Seems only fair but I'm not sure.

If anybody out there knows the ins and outs of Raceview please weigh in. ...
A bunch of years ago, I watched a strange program schedule races. One of the "raceview" comments elsewhere on Derby Talk reminded me of that program. That program was, like double elimination, adaptive. That is to say, who races whom depends on the results of prior heats. Its strategy appeared to be to differentiate between closely matched opponents. The result was a peculiar imbalance in lane occupancy and opponent selection. But I did not record the details of the heat assignments, nor did I see the name of the software. So what I just said may not be applicable.

Your question regarding "only 1 or 2 guys on the second table" ... if you are confident that #2 was among the finalists (that got promoted to table 2) and that the pairings on table 2 were equitable, then yes, I'd say you were right about who was awarded 2nd place.

If raceview is, indeed, "adaptive", then I would need to see several of your actual races details ... pairings and results from each "heat" and what the program determined to be the final ranking. Then I would need to see how you make use of the table 1 results and the table 2 results.

When you "race" ...
Does everyone start racing on table 1?
When and how does anyone start racing on table 2?
How do you determine who races on table 2?
How many?


Stan
"If it's not for the boys, it's for the birds!"

Rob Shafer
Apprentice
Apprentice
Posts: 11
Joined: Tue May 12, 2015 5:46 am
Location: Phoenixville,PA

Re: How's it work mathematically ?

Post by Rob Shafer » Wed May 13, 2015 3:19 pm

We haven't done this yet. To date we've only used one arena (four lanes) so ideally we should make a bigger dia arena with 6 -8 "lanes/gates) about 48" dia . Its not really practical I mean it can be done but we have a prefab oak pub table top 38" 2" thick that we want to use. So it's easy half the work is done. ANYWAY think of it like you had two four lane tracks and the computer thinks its one 8 lane track. COOL if you placed based on finish times not position. But we can't do that has to be by finish position without you actually physically competing against everyone in the heat.

So My idea is we tell the software there are 8 lanes it assigns the lanes and we award ties for finish position, this all seems fair IF the software randomly assigns lanes/arena and opponents and everybody has equal chance of being matched against each other, on an equal amounts of battles (races) and equal chances to get a race where its you and one other or you alone. Im still back to what would be fair. And the more I think about this I'm thinking we run two separate races at the same time with two computers then get 4 finalist from each batch and combine them in a third and separate battle. UNLESS one of you guys thinks this works mathematically to be fair



User avatar
Stan Pope
Pine Head Legend
Pine Head Legend
Posts: 6888
Joined: Sat Jul 05, 2003 7:01 pm
Location: Morton, Illinois
Contact:

Re: How's it work mathematically ?

Post by Stan Pope » Wed May 13, 2015 8:49 pm

re treating two 4/ arenas as one 8/ arena, assuming that you use your current software and rank by ties:

I think that this makes too many assumptions.

First, there is the assumption that if entrants a, b, c, and d finished 1, 2, 3, 4 in arena A and entrants h, i, j, and k finished 1, 2, 3, 4 in arena B that had a, i, c, and d been matched in arena A that i would have finished 2nd. In fact, i might have finished 1st or, perhaps, even 4th! Competition matters!

There is also the assumption of transitivity, i.e. that assuming the result above and a subsequent heat showed that d and h finished 1, 2 in a heat, then had c and i been matched c would beat i. I believe that raceview gives considerable weight to transitivity. Joining the two arenas compounds that error since it would view i as inferior to a and b as inferior to h, and therefore, the elimination points accumulate more rapidly without justification. For instance, d and k were beaten by 6 other entrants, c and j were beaten by 4 other entrants, etc.

From my direct experience with an adaptive competition scheme (quintuple elimination racing 3 at a time), we have used two tracks when two successive races were independent of each other, i.e. the result of neither would affect the pairing in the other. For example if I had a group of 11 competitors with equivalent records, I could randomly match those competitors in groups of 3, 3, 3, and 2 and run those heats in any order without affecting the outcome of the process. In practice, we would start the first group of 3 on track A, the 2nd group of 3 on track B, the 3rd group of 3 on whichever track was next available, etc. For most of the competition, heats were completed at twice the rate that could be accomplished on one track. I don't think raceview would allow you this degree of latitude, although it has information internal that would allow it to do so.

The above observations assume that I have correctly identified raceview and its method of operation. When/if I get some actual examples of competitions, I will know for sure.
Last edited by Stan Pope on Thu May 14, 2015 11:47 am, edited 1 time in total.
Reason: Correct "race" to "rate" in next to last paragraph.


Stan
"If it's not for the boys, it's for the birds!"

Rob Shafer
Apprentice
Apprentice
Posts: 11
Joined: Tue May 12, 2015 5:46 am
Location: Phoenixville,PA

Re: How's it work mathematically ?

Post by Rob Shafer » Sat May 16, 2015 6:18 am

ok I follow, sorry there's no data to give you since we haven't run it. I might be able to dig up data on pervious single arena outcomes, but truthfully I don't want to take up your time with it. Ok so I know you've heard this one before. (A quote "bad lane") If the process ie: lane choice therefore arena choice and who you competing against in that arena is flawed,,,, Since its flawed for everyone is that fair to all? My biggest concern is that later as points build some people may not be matched against each other or may not have equal times on different arenas. Truthfully its a bunch of old guys spinning tops, but they want to bang into everybody at least a couple of times. lol

Thanks for you help, I know/knew it wasn't a simple thing and was hoping to hit somebody with raceview knowledge. I think I'll run two sets of races and then a final 3rd with the winners. nothing worse than a bunch of grumpy old men locked up a dusty woodshop for 4 hours. :O

Rob



User avatar
Stan Pope
Pine Head Legend
Pine Head Legend
Posts: 6888
Joined: Sat Jul 05, 2003 7:01 pm
Location: Morton, Illinois
Contact:

Re: How's it work mathematically ?

Post by Stan Pope » Sat May 16, 2015 9:09 am

Before you write this off ...

1. Tell me the usual number of competitors (tops) to compete and the number range you expect. From that I'll tell you some specific suggestions.

2. The data does not have to be from real competitions. You can make usable data for me as follows:
a. list typical competitors, prime your race manager software with those.
b. decide on a ranking of those competitors.
c. run the software, assigning heat finish according to the ranking set in step b and recording the race (except about every 5th or 6th heat roll dice ... or some other random process ... to determine heat finish order).

Change the ranking in step b and rerun step c


Stan
"If it's not for the boys, it's for the birds!"

Rob Shafer
Apprentice
Apprentice
Posts: 11
Joined: Tue May 12, 2015 5:46 am
Location: Phoenixville,PA

Re: How's it work mathematically ?

Post by Rob Shafer » Fri Jun 05, 2015 6:02 am

I'm not sure I follow at B What do you mean "assign their ranking" then run race.
I figure on having about 35-40 members participating.
For this sample how many do you need run.

The names are entered into a delimited file the software assigns lanes, the race is run and finish position is entered. After everyone has gone once it starts over without a visible method of "pairing" people, and so on until the end of "heats" Final race starts fresh with 4



User avatar
Stan Pope
Pine Head Legend
Pine Head Legend
Posts: 6888
Joined: Sat Jul 05, 2003 7:01 pm
Location: Morton, Illinois
Contact:

Re: How's it work mathematically ?

Post by Stan Pope » Fri Jun 05, 2015 8:58 am

Rob Shafer wrote:I'm not sure I follow at B What do you mean "assign their ranking" then run race.
I figure on having about 35-40 members participating.
For this sample how many do you need run.

The names are entered into a delimited file the software assigns lanes, the race is run and finish position is entered. After everyone has gone once it starts over without a visible method of "pairing" people, and so on until the end of "heats" Final race starts fresh with 4
Since I am asking you to run some "simulations" without actual runs on the "track", you need a way to decide how the competitors finish in each "heat." Suppose the competitors are named "A", "B", "C", ..., "Z", "AA", "BB", ..., "NN". Then assign a ranking number (in range 1 through 40) to each such that when they race, the heat result is determined according to their ranking number, "lower is faster."


Stan
"If it's not for the boys, it's for the birds!"

User avatar
Stan Pope
Pine Head Legend
Pine Head Legend
Posts: 6888
Joined: Sat Jul 05, 2003 7:01 pm
Location: Morton, Illinois
Contact:

Re: How's it work mathematically ?

Post by Stan Pope » Fri Jun 05, 2015 11:42 pm

This may help you execute each simulated competition: When you prepare the list of competitor names, add the rank to the end of the name. Then you won't have to look up their ranks when recording the heat results.

For each list of competitors and ranks, you can run the schedule once deciding heat results by "losest rank number is faster". Then run the schedule again deciding heat results by "highest rank number is faster".

You can also add some randomness as follows:
For each heat, roll a pair of dice. If one of the dice is a 6 and the other is 1 through 3, then adjust as follows:
Second die is 1: swap 1st and 2nd.
Second die is 2: swap 2nd and 3rd.
Second die is 3: swap 3rd and 4th.
Otherwise record according to the rank number.

Okay, there are 3 race simulation plans to run and pass me results. That should be enough to eek out the algorithm, I think.

If you wish, you can cut down the number of competitors to 25 or 30. Of course, I need to see each heat's competitors and results, and final results.


Stan
"If it's not for the boys, it's for the birds!"

Post Reply