Train your killdude? Are you sure?

[Hurkyl]

I've been thinking about the usual starting advice that one should initially focus on training up your method of killing things until it's up to some level of adequacy before looking at training up defensive skills.

TL;DR: initial investigation suggests that for melee combat, you shouldn't let your defenses lag behind your weapon skill, at least if you have worthwhile ones to train. (e.g. a melee Sp background probably shouldn't train Armour for a long time!)

I'm not really sure that's actually a good idea.

As an experiment, I took a HuBe to fix the starting package, and was curious how the first few skill levels affect one's ability to fight against a jackal. (without berserking)

A HuBe starts with 3 skill in Maces & Flails. Based on fsim figures, training up to 4 skill increases your average damage dealing abilities by around 1.0652%, and costs 300 skill points; the score* of this method of training is thus 0.912.

(fsim suggests that average damage grows like 3.5 + 0.05 * skill, accuracy is like (55 + 0.5 * skill)%, and delay is 1.4 - 0.5 * skill)

Training dodging up to 2.8 is worth an extra point of EV. This reduces the jackal's accuracy from 43% to 39%, and costs 216 skill points. The score is 1.963. A much better return on your investment.

Okay, maybe that's because the first points are so cheap and we were already part way to the next point of EV. The next points of EV come at 4.5 and 6.1 skill, costing 604 and 1108 skill points, and reduce accuracy to 35% and 32% respectively. The score of these trainings are 0.778 and 0.351.

But M&F training gets expensive rapidly too; going from 4 to 5 skill only has a score of 0.562, and 5 to 6 has a score of 0.384.

The effect of training fighting up to 4 is a bit complicated, but if I assume it raises your HP to 108.8% of base from 106.6% of base, then its score is 0.296. So we're not at the point yet where it's worth training fighting, but it's not that far behind.

*: Score is log(improvement) * 10000 / skill point cost. The logarithm is because we need to convert a multiplicative benefit to an additive measure. Indicated figures are not accurate to the full precision displayed

[daggaz]

There are WAY too many factors to extrapolate that kind of message from a simple FSIM with a jackal. You point out just a few of them.

It would be far better to play, say 100 runs each to XL 10, changing only your skilling pattern. For simplicity, 100 runs at 100% melee, and 100 runs at 50% melee, 50% main defensive skill. Then look at the total results, giving you a somewhat averaged comparison, (for a given species/background, best not to take any gods on this, nor should one pick up and use anything but mundane gear). Thats a somewhat more objective take on what the tavern has been doing and where the rule of thumb comes from, and said rule of thumb does work out pretty good it would seem.

I would agree, however, that it can quite often be advantageous to take a few levels of a secondary/defensive skill if that skill is cheap, rather than letting it sit at 0. I also rarely go straight to min-delay unless its SB, without putting some points into secondaries, and quite a few other tavernites do the same.

[Hurkyl]

Just realized that the same methodology should work for comparing weapon skill vs dodging for dealing with an adder (which is what I had originally wanted to use as the example monster), since even with poison, the relevant quantity is still how much damage you're doing versus how often you get hit. The numbers are different, but the relative ordering is similar.

[partial]

melee fighters can afford to train defenses earlier, as there are fewer offensive skills for melee bashers to focus on as opposed to casters (fighting+wep skill vs. all the various spell schools)

[Hurkyl]

melee fighters can afford to train defenses earlier, as there are fewer offensive skills for melee bashers to focus on as opposed to casters (fighting+wep skill vs. all the various spell schools)

If you can't afford to train defenses earlier and training defenses earlier is better than later, then you're in a pretty bad spot.

A more accurate point of view, IMO, is that melee fighters want to train defenses earlier; ranged attackers don't get as much use out of defensive skills because they're not attacked often.

[MrPlanck]

In early game I almost alway focus weapon but keep defenses on. Works pretty well for me.

[sanka]

Do not neglect defenses, yes, but there are dangerous enemies in the early game where 1-2 point of EV does not matter as much as against a jackal, but you still want to have strong offense against them.