Wild Dog wrote:
I think it's a question of changing the value for each race category (WT, .hc, .1 etc.) so the difference is bigger. That way the Tour for example will count even more and riders such as Sagan and Valverde will not be the best in sprints and mountains respectively.
That is implemented;
If you look in the sheet at the points table, you will find that
top 5 in 1 day race will get;
1.WT1
1.WT2
1.HC
1.1
1
280
220
140
80
2
170
132
84
48
3
145
112
72
40
4
125
96
62
36
5
112
86
54
32
top 5 stages in tour of value get;
GT1s
GT2s
2.WT1s
2.WT2s
2.HCs
2.1s
140
80
46
35
28
20
84
48
28
21
16
12
72
40
24
18
14
10
62
36
20
15
12
8
54
32
18
13
10
7
Of course the points are not handed more then just top 5, but to compare.
So winning a stage in the tour is the same as winning a 1-day race of 1.HC category.
@Paul23; you are right, being top 2 to 10 every sprint in the Tour de France doesn't mean that you are better than Kittel who finish first a number of times... but I would like to think that when you're playing PCM with a number of sprinters that have the possibility to win, than it would be OK that Sagan is 1 of them. but as stated before;
Of course I can adjust this while copying and pasting it into a database. Give Kittel 84, Cavendish 83 and the rest 82...
Mind you I'm not defending my algorithm, I know it is far from finished and has a number of improvements and enhancements that could be made. But I would like to base them on more then just a couple of individual results that could be better.
I'm currently working on the algorithm that for the fighter stat, but still struggling on how to proceed. Will post some results later on to get some feedback.
also tried to calculate the fighter stat... this is not something I can calculate based on the number of hours spend in a breakaway... (if anybody has such a database or a link to it, than feel free to share )
But the result is somewhat satisfying, see for yourself. Anybody missing?
I knew that name would come up... I tried to get the algorithm to push him upwards, but then all other riders would get messed up stats.
CyclistName
FLA
MOU
HIL
TT
PRL
COB
SPR
ACC
DHL
FTR
STA
RES
REC
POP
VOIGT Jens
60
62
64
57
68
50
56
61
73
74
58
69
65
69
He's "only" 74 fighter... his popularity is also off in my opinion, but he doesn't get good (big) results, that's why the popularity calculation is so low.
Edited by TankNL on 11-08-2014 18:08
Lachi wrote:
I think it is a good system even if random riders like Voight or Sagan do not fit. It is easier to fix certain riders than to create a whole DB.
2 thoughts strike me looking through the most recent set of data.
the being that the sprint stats on a whole look high, highlighted by 7 guys being on the joint highest sprint and after looking at the raw points value of them I believe that the points to stat threshold for sprint might need adjusting a little as there's quite abit of variation in the points for those tied on 83. The lowest is Cavendish on 4690.17 points and the highest is Sagan on 6953.395 points.
Judging solely off the point value Sagan and Greipel should probably be joint highest, then Kristoff, Kittel then Bouhani and Degenkolb. Then Cavendish who isn't that much higher then Demare and they could be on the same level.
The second thing is that Valverde tends to have alot of racedays so he has a lot more chances to score points. I'm not sure how many more he has compared to Froome, Contador and the rest but I'd imagine it has had some impact on his points value and subsequently on his stats.
I was thinking it might produce a better outcome if there was a way of dividing the points by the umber of racedays as it should bring guys like valverde who race a lot and have a lot more chances to score points back to a more comparable level to others and would help avoid penalising anyone who has crashed and had to miss large or key parts of their season. For example Boonen missing the cobbled races in 2013 although it doesn't look like that's had an effect in the current stats.
Yes I know that could create a few more problems like Wiggins and Sorenson doing well in cobbles despite only taking part in 1 years worth instead of multiple seasons like the specialists but it should help the top guys anyway.
Although looking at the latest results Cobble might need stretching out in the same way, though I've not yet looked at their raw points value it still seems a little odd seeing 5 guy tied on the highest stat.
As announced, I've created a .cdb based on ASOdb2014.
Did nothing else but copy and past the stats from the Excel.
I only did a couple minor adjustments, just because I feel like that would become more realistic (like I said in this thread, its not about an individual, but about the Algorithms);
Mountain;
Valverde 83 -> 80
Contador 82 -> 83
Froome 82 -> 83
Ten Dam 77 -> 78 (because I'm Dutch )
Peraud 79 -> 80 (could be 81, but then Its solely on the TdF)
sammyt93 wrote:
2 thoughts strike me looking through the most recent set of data.
the being that the sprint stats on a whole look high, highlighted by 7 guys being on the joint highest sprint and after looking at the raw points value of them I believe that the points to stat threshold for sprint might need adjusting a little as there's quite abit of variation in the points for those tied on 83. The lowest is Cavendish on 4690.17 points and the highest is Sagan on 6953.395 points.
Judging solely off the point value Sagan and Greipel should probably be joint highest, then Kristoff, Kittel then Bouhani and Degenkolb. Then Cavendish who isn't that much higher then Demare and they could be on the same level.
The second thing is that Valverde tends to have alot of racedays so he has a lot more chances to score points. I'm not sure how many more he has compared to Froome, Contador and the rest but I'd imagine it has had some impact on his points value and subsequently on his stats.
I was thinking it might produce a better outcome if there was a way of dividing the points by the umber of racedays as it should bring guys like valverde who race a lot and have a lot more chances to score points back to a more comparable level to others and would help avoid penalising anyone who has crashed and had to miss large or key parts of their season. For example Boonen missing the cobbled races in 2013 although it doesn't look like that's had an effect in the current stats.
Yes I know that could create a few more problems like Wiggins and Sorenson doing well in cobbles despite only taking part in 1 years worth instead of multiple seasons like the specialists but it should help the top guys anyway.
Although looking at the latest results Cobble might need stretching out in the same way, though I've not yet looked at their raw points value it still seems a little odd seeing 5 guy tied on the highest stat.
Thanks Sammyt93, your input is very usefull... I actually had implemented an average calculated by number of race days, but got a lot of crazy results because of multiple cyclist doing good for 5 results (one 2.1 tour for instance) and would mess up all calculations. The other thing is, that injuries happen very often, whether we like it or not. So, missing out on a race goal for the season is big, but what about Jurgen Van den Broeck who was ill during the tour, or Mollema, how was ill the first week....
For me then it because interpretation again and I would like to keep looking at the results, find patterns and use and algorithm to calculate based on those patterns.
If you look at the result tab, then you will see that all results within 365 now count 5 times, all within 730 days count 3 times and the others 1 time. So that also might explain why Cavendish has less points, he has not scored a WT podium this year and his victories of last year count less than podium places of for instance Demare or Sagan of this year.
But if you could give me hints on how to improve the points distribution, than I would be more than happy to give it a go. After all, your idea about the domestics is what made the stats actually usable
sammyt93 wrote:
the being that the sprint stats on a whole look high, highlighted by 7 guys being on the joint highest sprint and after looking at the raw points value of them I believe that the points to stat threshold for sprint might need adjusting a little as there's quite abit of variation in the points for those tied on 83. The lowest is Cavendish on 4690.17 points and the highest is Sagan on 6953.395 points.
Judging solely off the point value Sagan and Greipel should probably be joint highest, then Kristoff, Kittel then Bouhani and Degenkolb. Then Cavendish who isn't that much higher then Demare and they could be on the same level.
The formula for calculating the stats is on the stats sheets.
I take the highest (actually second highest, because highest is the total of the column) scored points of the sprint column in the Pivot table.
Then I take the max points - 10 (in case of Sprint 83-10 = 73 and increase that by the points scored divided by (highest scored points divided by 15).
So 73 + pointsScored / (highestPointsScored/15) = SprintStat (fail-save is build-in, so if it comes above the max stat, then the max stat is used)
But only for somebody who has more points than highest scored points divided by 15. If this is not the case then I will do 50 plus 25 times points scored divided by (highest scored points divided by 15).
So 50 + 25 * pointsScored / (highestPointsScored/15) = SprintStat (impossible to go above 75)
I could use other algorithm's but it is a dilemma of not wanting to create to big of a gap between top and rest and giving enough momentum in order to make the results translate into a scale from 50 to the maximum stat.
TankNL wrote:
As announced, I've created a .cdb based on ASOdb2014.
Did nothing else but copy and past the stats from the Excel.
I only did a couple minor adjustments, just because I feel like that would become more realistic (like I said in this thread, its not about an individual, but about the Algorithms);
Mountain;
Valverde 83 -> 80
Contador 82 -> 83
Froome 82 -> 83
Ten Dam 77 -> 78 (because I'm Dutch )
Peraud 79 -> 80 (could be 81, but then Its solely on the TdF)
Here is the edited .cdb file; https://dl.dropbo...lStats.cdb
I hope a couple of you guys can help me with some test runs.
Forgot to give credits to OlegTinkov, who did a fantastic job with his database. Used it already for some time and really like it. That's why I used his db to experiment.
I love the Idea of using the multiplier for a decay of results, assuming you are able to keep this up to date will the results change bracket automatically?
Maybe adjusting the multiplier to 1, 1.5, 2.5 instead of 1,3,5 would help those that have missed a race goal for a season but I don't know what effect that would have on everyone else or on the stats in general.
sammyt93 wrote:
I love the Idea of using the multiplier for a decay of results, assuming you are able to keep this up to date will the results change bracket automatically?
Maybe adjusting the multiplier to 1, 1.5, 2.5 instead of 1,3,5 would help those that have missed a race goal for a season but I don't know what effect that would have on everyone else or on the stats in general.
the problem then is that Aru, Kelderman, Maijka won't be more than in the middle of the pack and guys like Scarponi and Basso will be more in the top, than the should be. But feel free to toy around with the numbers... that's why I do this "project" open source...
sammyt93 wrote:
I love the Idea of using the multiplier for a decay of results, assuming you are able to keep this up to date will the results change bracket automatically?
Maybe adjusting the multiplier to 1, 1.5, 2.5 instead of 1,3,5 would help those that have missed a race goal for a season but I don't know what effect that would have on everyone else or on the stats in general.
the problem then is that Aru, Kelderman, Maijka won't be more than in the middle of the pack and guys like Scarponi and Basso will be more in the top, than the should be. But feel free to toy around with the numbers... that's why I do this "project" open source...
what if domestique points had an inverse over time multiplier to individual points, that might help young riders have recently developed into leading roles like Kelderman and Majka, though however you do it you are still going to end up penalising riders that have turned pro since you start taking the results, e.g. the Yates brothers, Ewan etc.
now it looks quite good...
i would still give voigt 85 in FTR....but thats maybe because im german...
BTW: nice, that you've made some adjustments on the stats. now they look better than before...
you said, you make this project open source...so we could make some tests ourselfs and we could write you suggestions, or what do you mean by open source(yes, i know what open source is, but everyone handles it a bit different)
Paul23 wrote:
now it looks quite good...
i would still give voigt 85 in FTR....but thats maybe because im german...
BTW: nice, that you've made some adjustments on the stats. now they look better than before...
you said, you make this project open source...so we could make some tests ourselfs and we could write you suggestions, or what do you mean by open source(yes, i know what open source is, but everyone handles it a bit different)
Thanks Paul23, yes by open source I mean that I post all my source files with I use to create / generate the stats. I do that so that people can see and understand how the are generated and also to suggest or discuss topics on that matter. Sammyt93 has posted several good idea's and some of them are used in the model that I used to create the .cdb
I'm trying hard to appreciate you opinion, but can you explain yourself a little bit more? Did you do a test run? Are you replying on the results that OlegTinkov has posted? If so, what is average in you oppinion? Without that, you comment is rather useless, but that is just my honest opinion.
Edited by TankNL on 12-08-2014 03:52
I'm trying hard to appreciate you opinion, but can you explain yourself a little bit more? Did you do a test run? Are you replying on the results that OlegTinkov has posted? If so, what is average in you oppinion? Without that, you comment is rather useless, but that is just my honest opinion.
About Oleg results.
Chavanel only lost 3min to Nibali at stage with mountains and mountains.
im not have db, but im test here simulate stages and i see what results im have
)
