bssop97 Posted August 3, 2005 Share Posted August 3, 2005 Just curious... How do we get generated scores. What is the formula based off of? Quote Link to comment Share on other sites More sharing options...
Brass Ablaze Posted August 3, 2005 Share Posted August 3, 2005 I'll second that question. Pretty sure it has something to do with a slinky, a cat, and the number of green m&ms that are in a random bag. :P Quote Link to comment Share on other sites More sharing options...
sburstall Posted August 4, 2005 Share Posted August 4, 2005 (edited) Darn it!!!...you gave it away (I prefer working with Tabbys or Russian Blues over Siamese). :P No seriously. It took a while to develope it (more like an hour being bored at work). There are two parts to the generated score...if you're anti-math, you better go to a different topic. This will sound messy (but it actually isn't). :P Effect......Visual............Music Vis...Mus...Perf..Clgd..Ens...Brss..Ens...P.Fld.P.Ens.Score 14.70 14.10 14.00 14.20 14.40 14.10 14.50 14.80 14.00 71.600 Part 1) I line fit each caption and the final score. That's done with this spreadsheet command =TREND(B10:B28,$A10:$A28,'JR. SCORES-05'!$C1). B10:B28 are the captions ranking. These change from row to row. The $A10:$A28 are the dates of the show. The $ freezes the cell A so that doesn't change as I copy and paste the command across the whole row. 'JR. SCORES-05'!$C1 is the command to bring in the date from the worksheet called 'JR. SCORES-05' So any generated scores in the "live season" are actually exterpolations from the caption scores (exterpolated means outside the data set, interpolated mean inside the data set...which is what the "off-line season" does). The line-fitted Effect Visual for Pioneer as of yesterday was 15.22. Part 2) This comes from the previous year's DCI Finals caption scores. This adds that little bit of zing to the generated score. Above is the caption and Finals score from 2004 DCI Finals (thier last show of the year). I figured out the percentage that each caption in relations to their final score. So for Effect Visual, the percentage is 14.7/71.6 = .205. Now using the line fitted final score from Part 1), in this case, Pioneer's line-fitted final score from yesterday was 74.043. Then the Percentage caption score is .205*74.043 = 15.20(yes, it's higher than the real caption score). I then take a weighed average of the line-fitted caption score with the Percentage caption score. It's weighed (2*Line-Fitted + Percentage)/3 or (2*15.22+15.20)/3=15.1. I give more weight to the line-fitted caption score than the Percentage. It's not true in all cases that generated scores go up. They also come down. I hope I didn't lose anyone. I tried to explain it as best as possible short of giving away the spreadsheets. Edited August 4, 2005 by sburstall Quote Link to comment Share on other sites More sharing options...
Brass Ablaze Posted August 4, 2005 Share Posted August 4, 2005 Well, I read this like 5 times and I'd say I understand about a third of what you are saying. :) So basically it takes their last score and predicts a trend. So why are you using last years finals scores again? Isn't that going to inflate the generated scores more often than not? Quote Link to comment Share on other sites More sharing options...
Brass Ablaze Posted August 4, 2005 Share Posted August 4, 2005 Why don't you just take the average of the last three differences in that caption and add it to the last one? So for instance here is the spreadsheet: ...........A......B.......C.......D.........E 1.......................................GENERATED 2........7-Jul..8-Jul...9-Jul...10-Jul...11-Jul 3........15.11..15.20...15.22...15.60....15.76 4........................................... So in this example, the July 11th caption score is the generated one. The formula for cell E3 is =((D3-C3)+(C3-B3)+(B3-A3))/3+D3 Example #2: (same forumula) ...........A......B.......C.......D.........E 1.......................................GENERATED 2........7-Jul..8-Jul...9-Jul...10-Jul...11-Jul 3........16.05..16.12...15.98...16.09....16.10 4........................................... Example #3: (same forumula) ...........A......B.......C.......D.........E 1.......................................GENERATED 2........7-Jul..8-Jul...9-Jul...10-Jul...11-Jul 3........17.00..17.31...17.33...17.43....17.57 4........................................... Now here lies the problem... Lets use example 3. Lets say that on the 10th, its a judge that gives them an off the wall score that doesn't fit with the pattern. Lets say 16.50. In that case, the new generated score trends down to 16.33. That didn't sound good to me so I tweaked it a little... SO....I updated the formula to average the last 4 scores into it as well, so that one bum score wont throw it way off the wall. =((((D3-C3)+(C3-B3)+(B3-A3))/3+D3)+D3+C3+B3+A3)/5 Now instead of 16.33, the generated score would be 16.89, which is a little more realistic probably. That would make the generated score for example #1 be 15.38 and for example number 2 be 16.07. Those scores seem a little less dramatic to me, which is better. The point is, you dont want a corp's caption to suddenly change drastically based on the generated score. The generated score should as accurately as possible try to predict the most likely score for that caption based on the last few scores they have gotten in that caption. That's my take on it anyway. B) Quote Link to comment Share on other sites More sharing options...
sburstall Posted August 4, 2005 Share Posted August 4, 2005 There's nothing "realistic" about scores. If you've seen the spreadsheet, the captions fluxuate up to a point depending on the judge and day. Plus I don't scores to continually go up. That's even less realistic. (I knew trying to explain statistics was going to be hard). Line-Fitting creates a line with a given set of data. And yes, the dates in which the scores are done are taken into account. As you can see is that line fitting does increase, but I temper it with a percentage score so it won't ALWAYS go increase. Quote Link to comment Share on other sites More sharing options...
Brass Ablaze Posted August 4, 2005 Share Posted August 4, 2005 well it's not really "increasing". If you just went by this it would be fine because lets take, for instance, the day that it was a 14.70. If the following day's score was generated based just on this, it'd be something like 13.75 it looks like. So that would actually be accurate according to your line but not an "increase" like you said. Quote Link to comment Share on other sites More sharing options...
Galen Posted December 7, 2006 Share Posted December 7, 2006 Darn it!!!...you gave it away (I prefer working with Tabbys or Russian Blues over Siamese). :P No seriously. It took a while to develope it (more like an hour being bored at work). There are two parts to the generated score...if you're anti-math, you better go to a different topic. This will sound messy (but it actually isn't). :P Effect......Visual............Music Vis...Mus...Perf..Clgd..Ens...Brss..Ens...P.Fld.P.Ens.Score 14.70 14.10 14.00 14.20 14.40 14.10 14.50 14.80 14.00 71.600 Part 1) I line fit each caption and the final score. That's done with this spreadsheet command =TREND(B10:B28,$A10:$A28,'JR. SCORES-05'!$C1). B10:B28 are the captions ranking. These change from row to row. The $A10:$A28 are the dates of the show. The $ freezes the cell A so that doesn't change as I copy and paste the command across the whole row. 'JR. SCORES-05'!$C1 is the command to bring in the date from the worksheet called 'JR. SCORES-05' So any generated scores in the "live season" are actually exterpolations from the caption scores (exterpolated means outside the data set, interpolated mean inside the data set...which is what the "off-line season" does). The line-fitted Effect Visual for Pioneer as of yesterday was 15.22. Part 2) This comes from the previous year's DCI Finals caption scores. This adds that little bit of zing to the generated score. Above is the caption and Finals score from 2004 DCI Finals (thier last show of the year). I figured out the percentage that each caption in relations to their final score. So for Effect Visual, the percentage is 14.7/71.6 = .205. Now using the line fitted final score from Part 1), in this case, Pioneer's line-fitted final score from yesterday was 74.043. Then the Percentage caption score is .205*74.043 = 15.20(yes, it's higher than the real caption score). I then take a weighed average of the line-fitted caption score with the Percentage caption score. It's weighed (2*Line-Fitted + Percentage)/3 or (2*15.22+15.20)/3=15.1. I give more weight to the line-fitted caption score than the Percentage. It's not true in all cases that generated scores go up. They also come down. I hope I didn't lose anyone. I tried to explain it as best as possible short of giving away the spreadsheets. :sshh: Quote Link to comment Share on other sites More sharing options...
sburstall Posted December 7, 2006 Share Posted December 7, 2006 Hush you....<_< Quote Link to comment Share on other sites More sharing options...
StarOrg Posted December 7, 2006 Share Posted December 7, 2006 I guess I have a question now for clarification. If said corp's caption is being used and that corps has a real score on a matching date for all 3 years; 2001, 2003, 2005. Will that score be the average of those? OR are you still really generating a score based on the methods above therefore generating scores everyday? Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.