scerruti wrote:Assume two hypothetical drivers. D1 goes straight, D2 turns. The only way that D2 beats D1 is if D2 arrives prior to Ta and Tx < (Tc+Td).

I have a few problems with this.

1) What you are essentially saying is that D2 will arrive at the light Tx later than D1. But presuming that at T0 the light could be at any point in the cycle, the actual point in the cycle when they reach the light will be random. As long as Td>Tb then, in the absence of traffic, on average it is going to take D2 less time to clear the intersection than D1. The question is whether the difference in the times to clear the intersection is greater than Tx or not.

2) I don't think Waze is calculating the probability that for any T0 in the light cycle D2 will be faster than D1. Rather waze seems to be averaging the times. So even if D1 is faster more often, if D2 is so much faster when it is faster that the average for D2 is faster, then I think waze chooses D2.

3) None of this really matters in the hypothetical no traffic situation. I agree that that if Ta is much greater than Tb, Tc and Td, then the average time for D2 to clear the intersection is unlikely to be Tx faster than the average time for D1 to clear intersection. But all this changes if many people are turning left from Oceanside to Crouch and there is little traffic on Crouch. Then you may bypass the queue by going around the block. Assuming that waze has valid data for Tm, it seems to me that waze could be detecting traffic waiting to turn left.

4) I still suspect that waze is underestimating Tm and that if D2 is driven a few times, waze will learn that Tm is greater and stop routing this way.