I do this, and I'm pretty glad to see I'm not the only one.
Applying query tuning techniques to my daily commute saves me about 10 minutes of driving each day. I recently realized that freeway lanes have similar performance characteristics as database tables do - the more transactional the lane, the more likely it's going to have a backlog. So, instead of driving in the fast lane, the one immediately to the right of it is usually faster. My guess is that this happens because of people swerving in and out of the HOV lane illegally in an attempt to save a few minutes or whatever. Avoiding the furthest 2 right lanes is important when you can as well as people tend to merge twice immediately after entering the freeway (this is on I-10 in Phoenix, which is between 4 and 7 lanes across each way, depending on what part of town you're in).
Knowing which lanes contain forced merges allows for planning ahead as well, since you know that you're going to have to deal with double the traffic in a rate-constricted channel.
There cannot be more lane changes into and/or out of the left most lane than the 2nd to left most lane (assuming no left exits) because all lane changes go through it. Furthermore it is fairly easy to get a sense of the best lane by just paying attention.
Yes there can be, because people who hop in and out of the HOV lane don't traverse all lanes every time they do it.
For example, Alice gets on the freeway and merges 4 times (after the initial merge) to get to the "fast lane". For ease of explanation, the lanes will be referred to like this:
| HOV | 5 | 4 | 3 | 2 | 1 | Onramp |
(So lane #5 is the fast lane in this example)
Traffic starts slowing down, so she merges from lane 5 to the HOV lane in order to pass a few cars. She's in this lane illegally so she doesn't remain in this lane. After a few passing a few cars, she merges back into lane 5. She repeats this 3 more times before she gets off the freeway.
So, in this limited example, this is the "transaction count" of Alice's lane changes:
Onramp <-> Lane 1 : 2
Lane 1 <-> Lane 2 : 2
Lane 2 <-> Lane 3 : 2
Lane 3 <-> Lane 4 : 2
Lane 4 <-> Lane 5 : 2
Lane 5 <-> HOV : 8
The problem isn't people just getting over to the fast lane, staying in it and leaving, the problem is the number of transactions that the HOV <-> Fast lane pairing have due to it being relatively empty even during crowded freeway conditions.
Similarly, lanes 1 and 2 have higher transactions due to people getting on the freeway for a mile or two and exiting shortly thereafter. This may be limited to AZ freeways since the exits tend to be regularly spaced about a mile apart. The other problem is that there are quite a few people who will merge on to lane 1 going 45MPH, which causes many drivers to pass them in Lane 2.
Chronic lane changers seem to use all lanes roughly equally, so their habits mostly cancel out.
Another thing that may be different depending on where you live is that AZ allows entry and exit from the HOV lane at any time. It's not like CA where there are specific entries and exits to the HOV lanes.
By "fast lane", I believe icey is refering to lane 5, not HOV. He is saying lane 4 is better than lane 5 because of all the people switching between land five and HOV. He's not saying that lane 5 is better than HOV.
In the UK people in the 'fast' lane (over taking on the inside is illegal here). Will generally leave less space between cars, so that lane is more likely to suffer from a cascade of over-breaking causing a jam.