v3.3.0 Evaluate mode

v3.3.0 'evaluate' mode

Evaluate mode sims for each turn which of the options to play is the best and plays it, like 'flexible'. Instead of simming all options to the end a score for the whole board is calculated to approximate the situation.

In detail:

Every turn you have 3 (or less) cards to choose from to play. Evaluate will simulate what happens when you play one of those 3 cards. [A simulation in the simulation] Then for each card the resulting board is evalued (after one turn by the player and one by the enemy). The card with the highest score gets played.

Pro:

• Not as slow as flexible, but still significantly better than random (~8% better in 10x time vs random)
• More realistic simulation, due to dynamic order instead of a static/random order.
• Includes (multiple-)dynamic combos in playstyle.
• Simming enemies with evaluate order represents a more realistic random defense vs a manual offense.

Con:

• The Algorithm (eval-function) for judging the board has it's issues, eg. fine-tuned scaling of some skill values
• Some good decks might be avoided since they don't show a the board dominance to appeal to the eval-function

Options:

• 'eval'/'evaluate' : instead of ordered/random-order
• 'enemy:evaluate' : instead of ordered/random-order
• 'eval-iter X' : sets the number of iterations to X for the simulation which card should be played (default=8)
• 'eval-turn X' : sets the number of turns to use the evaluate method, after X turns random is used (default=8)

Notes:

• The output is not ordered, that means you need to figure out how to play this deck yourself.
• Before simming vs a gauntlet, better get a feeling for it vs a single deck.
• Values for normal iteration and (default-)eval-iter are not optimized (yet), just test and see how tuo responds.

If you have ideas how to easily/quickly and effective judge/evalue the board I'm happy to hear about new approaches.

v3.3.0 'evaluate2' mode (cont.)

'eval2' mode is like 'eval' mode, but one more simulation of 1 (atm hardcoded) iteration follows in order to provide better plays:

1. simulation of the game
2. evaluation of the 3 available cards to play
3. evaluation of the 3 available cards to play after playing (2.)
4. Play the card from (2.) with the best score in (3.)

Pro:

• Scores are close to 'flexible' scores, but there is a gab.
• Quicker optimation of a bad start deck than starting with 'flex' (~twice as fast)

Options:

• Uses the parmeters from 'eval'