This strategy is Brourd’s idea. I put in the constraints.
If a shape appears solved in the EteRNA interface, it should have no unbonded pairs in the eterna energy model. No red spots in the target shape box.
Our robots made a 100% design in Shape test lab. This design also had one unbonded pair. So the Ensemble energy model did best, though it looked unsolved in EteRNA energy model.
We know the current EteRNA energy model is flawed. So we don’t want a strategy that downvotes designs, like the Ensemble strategy winner in Things to test lab. Most of the latest ensemble winners have an unbonded pair, according to the EteRNA energy model. We could limit our ensemble strategy unnessesary.
PATTERNS FROM PAST LABS
The lowest scoring ensemble designs have more unbonded pairs than the higher scoring. Not always, but the tendency is there.
The best Kudzu design by the EteRNA ensemble strategy (90%) had three unbonded pairs.
One of them should disappear after Brourd’s 1-1 loop strategy. So the best of the ensemble strategies from Kudzu lab, would then have had two unbonded pairs.
In Water striders all of the ensamble designs are full of unbonded pairs and unsolved.
In Things to test, the ensemble strategy is well behaved. Not one pair of unbonded. Same with A tilted running man.
I think the ensemble strategy were set to follow EteRNA energy model back then. So bonds on our bot have been loosened and maybe a bit too much.
There are two reasons why there is unbonded pairs. One: Flaws in the ensemble strategy. Two: Flaws in the eterna energy model. We want to catch the flaws but allow the brilliance of our little bot, when he outsmarts the eterna energy model.
Two unbonded pairs should be allowed to slip. But three should create worries.
THE STRATEGY
Allow two unbonded pairs (in the eterna energy model) for ensemble designs without penalty, as most of the late ensemble strategies have one unbonded pair, the best designs too.
For each extra unbonded pair more than two, give -0.5 and penalize exponentially.
I asked Jerry if I was allowed to penalize like this. He said that they could program: ”the first point would be -.5, then -1, then -2, -4, -16, and so on”
I’m aware that the calculated design score gets rounded of.
Thanks to Jerryfu, EteRNA programmer, for answering questions in relation to the constraints of the strategy.