I was thinking about how glitchy the AI can be with a small peloton and decided to do some testing and in the end decided to share my results.
I began the tour with certain Startlists and played stages 2,3.4,5, 10, 11 and 12 (simulating the stages in between) to test out the Startlists on different terrains.
Normally when you have a 20- rider AI peloton the break will ride away to the win, but I decided to experiment with some different situations of this.
First I took the best climber (then puncheur, then time trialist as tie breakers) of each world tour team plus MTN and Cofidis and added Duarte as a control rider to build a 20 rider field I then began the Tour de France:
Stage 1 (ITT): Simulated (won by Chris Froome)
Stage 2 (flat): they did nothing all race, Duarte had a heart rate of roughly 69 for most of it, until the final 2 kilometres where Pozzovivo won the sprint.
Stage 3 (hilly): nothing again until the final 12km this time, where Martin attacked, over every hill they would attack each other, then slow down on the descent, finally on the penultimate climb Uran, Yates, Pozzovivo and Froome got away with Uran winning the sprint.
Stage 4 (Cobbles): Nothing until 8km when Mollema attacked, Kelderman, Nibali and Costa chased but got caught when Contador tried to attack, which split the pack in half (Pozzovivo behind) while Mollema won solo 19 seconds ahead of the front split of the peloton.
Stage 5 (flat - Undulating): Nothing until the sprint, won by Uran.
Stage 6 (hilly): Simulated won by Rodriguez
Stage 7 (flat): Simulated won by Costa
Stage 8 (hilly): Simulated won by Costa
Stage 9 (TTT) Simulated won by Froome at this point Froome had a 1'05 advantage on Uran and a 1'14 advantage on Contador with Costa leading mountains and points (by one point each)
Stage 10 (Mountain - Undulating with a tough climb at the end): Nothing until the start of the climb (16km) where Pozzovivo, Quintana, Navarro and Yates attacked although Nibali and Contador chased to catch them with only Frank and Meintjies dropped. The response was for them all to attack again and Contador and Nibali attacked to catch them followed by Mollema and Rodriguez just behind. Froome, Uran, Pinot and Costa dropped all others in the chase but only caught Navarro and almost Rodriguez while Quintana won the front sprint and Contador in second moved within 25 seconds of yellow.
Stage 11 (Mountain - Category 1 climb and Hors category before am easier finishing climb): Attacks from the favourites on the Hors category climb (50km to go) nine riders got away although it did come back together this inspired Pozzovivo and Quintana to go with Contador, Froome, Rodriguez , Pinot and Mollema chasing a few attacks later Froome, Quintana and Contador topped the climb together but those seven mentioned minus Mollema came together on the descent with Nibali and Mollema chasing unfortunately they sat up on the descent and it mostly cam back together. The final climb had lots of attacks although the group wouldn't split until Quintana got away with 3km to go. He took the stage while Contador got a gap behind him and moved to 19 seconds off the lead.
Stage 12 (Mountains - Cat. 2, Cat. 1, Cat. 1, Cat.1): nothing until 50km to go on the penultimate climb (3km from the summit) when Yates and Barguil unsuccessfully attacked, at the beginning of the final climb (15km to go) 7 riders went including 4th in GC (1'56 down) Costa. Although they weren't let away for long plenty of attacks followed and Contador, Quintana and Froome marked each other while Nibaki tried to hole his third place. Contador got away with 7km left and both Froome and Quintana working together kept him at 25 seconds away. Contador won the stage and GC ahead of a five man group.
Conclusion: with the top climbers the AI really isn't that bad on mountain stages and the result was quite exciting with Contador taking the GC on the final stage, despite sitting up on descent they do have exciting finishes. Unfortunately they wait until the end of other races although it's not too bad with races acted out somewhat realistically just with little attacking.
Next I'm planning to try either one teammate for each rider or a more varied peloton, 20 top climbers, 20 top sprinters, 20 top cobblers, although I don't think you can put them on different teams, so I'll look into that.
Just something that interested me and I decided to share I'll likely update this with more tests in the near future.
Just one rider per team won't produce any breakaways, I guess. They'll all be counted as leaders with the order to do nothing except win the stage, which produces these late attacks. I wonder if even two riders per team would be enough to trigger an early break.
Ripley wrote:
Just one rider per team won't produce any breakaways, I guess. They'll all be counted as leaders with the order to do nothing except win the stage, which produces these late attacks. I wonder if even two riders per team would be enough to trigger an early break.
Anyway, I've managed to make a small peloton work 90% realistically. by placing them in a large peloton of 50 rated riders.
I tested this on a 19km course which was flat for 14km before climbing the mur de Huy and finishing after a further 500m flat.
I created a peloton with 100 riders all with 50 stats and added on 4 riders: Valverde (84hl) , Gerrans (81hl), Pellizotti (77hl) and Kangert (76hl), they all raced realisticly each time I tested, they sat at the front of the peloton but didn't do any work while riders attacked and other teams chased, then when they neared the end they would attack each other, usually with Valverde winning (but one time Gerrans beat him in a sprint).
I tried again added 15 riders for a Swiss championship, where the favourites Albasini and Frank sprinted for the win with Zaugg and the IAM team chasing hard behind.
It's not a perfect solution, especially as it only works in single race mode (unless you plan to add 100 riders into your career each time the NCs come round ) but just thought I'd mention it.
