František and Sean, Thanks for thinking out loud. You've given me an idea for a new class of strategies, which I'd like to try out today (except Wednesday is always busy, and today is busier than usual, so it will have to be tomorrow.
Look at the current values (or maybe latest few values) in low and high.
Look back (to the beginning, or just a limited memory). Have I seen this configuration before?
Look at my own history (I hadn't found a use before for this). What did I do last time (all previous times)? How did that work out at the time?
There's quite a bit of room for experimentation. How to use my experience of other rounds, how much memory to use, what does “this configuration” mean? I think I need to coarse grain the configurations, maybe just “too many to pay better than Stable”, “Few enough to pay”.
I have been using the learning approach from Fogel and al, and it needs some other strategies to shake it up.
František and Sean, Thanks for thinking out loud. You've given me an idea for a new class of strategies, which I'd like to try out today (except Wednesday is always busy, and today is busier than usual, so it will have to be tomorrow.
Look at the current values (or maybe latest few values) in low and high.
Look back (to the beginning, or just a limited memory). Have I seen this configuration before?
Look at my own history (I hadn't found a use before for this). What did I do last time (all previous times)? How did that work out at the time?
There's quite a bit of room for experimentation. How to use my experience of other rounds, how much memory to use, what does “this configuration” mean? I think I need to coarse grain the configurations, maybe just “too many to pay better than Stable”, “Few enough to pay”.
I have been using the learning approach from Fogel and al, and it needs some other strategies to shake it up.