Short answer
I can be 75% sure that the lights are synchronized within 0.02 - 0.2 seconds of one another.
Background
Interesting question, but hard to answer! Because it has been unanswered for quite a while I will try to formulate an answer that may provide an approximate value for the minimal asynchrony perceivable.
The temporal resolution of the visual system can be quantified in a number of ways. As you are referring to relatively simple stimuli, the critical flicker fusion frequency is probably the most relevant. At a certain critical frequency, a flickering stimulus will appear as a continuous stimulus (Holcomb, 2009). For example, the good old CRT screens can sometimes seem to be flickering. The mains is 50 Hz (US) or 60 Hz (Europe, Australia), and indeed the critical flicker fusion frequency is 5 - 50 Hz, dependent on the lighting conditions (Kalloniatis & Luu). Hence, stimuli separated by 200 ms (low lighting) to 20 ms (high lighting conditions) can be temporally resolved.
I am aware that your question is different, as you ask whether two stimuli on the retina can be perceived as being asycnchronous. Flicker fusion testing is mainly done with flash stimuli, i.e., a single stimulus per image, not two. However, from the various measures of temporal resolution of the visual system that have been used to date, the critical flicker fusion has the most conservative value, in that it provides the top-value of temporal resolution. Higher-level processing is much slower and values of around 4 Hz are obtained. For example, when subjects are tested on a stimulus of dots that are white when moving right and black when moving left, and they change directions/colors at a certain frequency, it becomes hard to tell whether the color change and direction change are synchronized at alteration rates of ~5 Hz and higher (Kalloniatis & Luu).
As to your question what the percentage chance is that stimuli separate by 20 -200 ms are separable, I can say that in general, psychophysical procedures determine the threshold defined as halfway between chance and 100% correct. Assuming a typical yes/no tasK (are the traffic lights synchronous yes/no?), chance would be 50% (guessing), and threshold would thus be defined as halfway between 50% and 100%, i.e., 75% correct.
References
- Holcomb, Trends Cog Sci 2009; 13(5): 216-21
- Kalloniatis & Luu, WebVision, chapter "Temporal Resolution" 2007