On Free Retreat Attacks

This came up in our game last Friday: When a figure retreats from melee, does the opponent get a free back attack? Unlike most standard versions of D&D, I say "no" in my games.

I think that the germ of this rule is from Chainmail mass combat -- that a retreating unit loses a turn as they attempt to rally and can be attacked for free if desired (somewhat poorly having chosen to stand motionless for a minute with backs to the enemy)...

 ... and, echoing Monday's post, this wound up being carried through sympathetically to man-to-man combat in AD&D (DMG p. 70):

Serendipitously, I ran into a video today by Lindy Beige on YouTube on exactly this issue, dealing with both the real-world simulation and game-design balance aspects. We approve:


  1. Back Attacks (Attacks of Opportunity) are fiddly and discourage players from actually retreating/moving about the combat zone.
    In my 2e game I let Thieves do it, cuz they need something darnit, but thats it.

  2. Agreed. Thinking about, Lindy Beige explanation
    reminded when I was a kid and most of the fights are like melee skirmished where we hit someone and then run. Usually, the one being hit only had a chance to hit back if he was considerably larger or faster. I think you can even watch some tag (the child game) to get some insights about the topic...

  3. Having done a year or two of HEMA, I think it's rather silly to imagine that someone engaged in a sword fight will just turn around and expose their back before running. You'd point your sword in their direction and back off quickly, keeping that sword up.

    It might well be different in the case of a unit of fighters engaged in combat with another unit though.

    I'm happy to assume that you don't get a full movement while backing off, and if you're running though a group of enemies then a few people might get a free attack. Common sense says that if you move carefully you should be okay.

  4. yeah. I much prefer to allow initiative and/or turn sequence to represent the "free attack".

    If you win init, you get to hit the other guys before they run away. Or, get to runaway before you get hit.

    1. That's pretty much just where I am. I do constant round-the-table sequencing (roll once at start); if you flee, then naturally right before that you took an attack that you can't respond to.

  5. I would add that in chainmail, retreat and rout occur only on serious morale failures... And rout allows 1.5 moves back, meaning that follow-up for that free attack may not be possible depending on move speeds.

    Most morale failures result in simply moving back in good order.

    The morale calculation is involved but it appears at first glance that the back attack scenario only occurs 40% of the time at most. And light foot in half of those cases will rout well beyond easy followup. On the flip side, heavy foot routing can be easily followed up by light troops, although it is rarer for heavy foot to rout i n The first place.

    The middle ground I suppose could be to only allow a back attack 2/6 of the time. Perhaps allow 1/6 for those fleeing at faster move, 3/6 if fleeing while slower, and allow rogues to roll pick pockets to negate a flubbed check.

  6. I don't find his presentation convincing for man-to-man fighting. Consider that both combatants are constantly on the lookout for an opening to attack. If one opens himself up at all, the other is able to lunge forward and ignore his own safety a bit because he knows that his enemy is not taking a stab at him. The enemy is at least looking in the other direction once or twice, even if he holds his shield up and continues shaking his spear, because he needs to make sure he doesn't trip over some piece of combat debris or uneven terrain. Note that around 8:22 when he describes the "right way to flee", he turns his head away for a second. That head is getting stabbed. If he leans in to feint, then lean back and step away in hopes of a fade to turn and flee, that's way worse! Because him shifting his weight and turning his body takes some time, and the opponent's spear in-hand is relatively very fast. It seems like he's giving a lot of leeway to the fleeing man and assuming that the attacker will stand still like a doofus. Like I said, not convincing.

    In 1E I give someone doing a 1/3 MV "fighting withdrawal" (he only moves, gets no attacks) one attack from his enemy at no bonuses. It's not a back attack. But it is a freebie. Someone who does a full move escape gets attacked from behind.

    Totally separate from the arguable realism aspect, I like that melee is sticky. I like that missiles fired in may strike friends. I think it adds to the tactics of the fight, where the losing side must realize they are losing as far in advance as they can; poor player skill results in a "well I can't even flee now because he'll pop me and I'll die anyway" situation. A skirmisher attacks to the outside where he can focus more men against a weak facing of the enemy. Skirmishers pick their attack to minimize retaliation. You don't send your skirmishers to slide along the front line of enemy spearmen if you hope to have them survive. Else the skirmishers fire or hurl missiles while the enemy line must hold, which is outside the example.

    I agree that someone moving around within the melee doesn't trigger the parting attack, and someone falling back behind friendly lines doesn't trigger it for the reasons he gave.

    1. I will add that no man retreats backward as fast as his enemy can approach walking forward. Again assuming man-to-man combat: the withdrawing man must actually turn and flee if he hopes to actually escape, otherwise he is more "giving ground" and the fight continues with both of them shuffling along. If you plan to flee, it's because you really want this combat to end, not because you want to continue it somewhere else. That latter fighter is trying to reach his reinforcements, or some more advantageous position, etc. But if you're injured or your armor is tangled, and you're alone in the dark alley vs. one opponent, a fighting withdraw just means your lifeblood is a streak instead of a pool.

    2. Sticky melee as you describe would result in much higher casualty counts than we observe in real world battles. Typically real world armies take 10-30% losses... When they break mkrale, they run. If the victors are faster (cavalry, light foot) then the killing rate goes way up. In the historical world many forces have fled without the disastrous consequence that 100% back attacks at major bonuses to hit would suggest.

    3. That's why I specified man-to-man engagements. If you have a unit 4 ranks deep, and the whole unit flees, 3/4 of the unit will receive no back attacks. The 1/4 who do will, according to D&D, probably only be struck roughly 60% of the time (0-level vs. AC 6, but +4 to hit fleeing target, and +1 for no DEX bonus, +1 for bypassing shield) for hits that may not kill them. And, in a large-scale battle, if one unit breaks and flees, will the attacking unit truly pursue and make good on their advantage, plunging deep behind enemy lines and suffering withering flanking attacks?

      Instead they might not even straight-up flee. They might withdraw behind friendly lines in a more orderly fashion, meaning the attack the enemy unit gets (in my game) would be at standard chance to hit (30% for 0-levels stabbing AC6). The end result is not some rundown mopping-up, but a free round of attacks.

      Next, a unit that breaks and flees should change to open formation, or even split up into individual men who scatter. The pursuing unit must decide whether to likewise split up to continue pursuit attacks or else roll on as a unit and mop up whatever scattered men they meet. That alone should drop the retreat casualties to a minimum!

      And does the pursuing unit want to chase them down? Or is there a more pressing objective?

      Personally I'm fascinated with why real-world casualty rates in conflicts and attacks are so low. Someone will run through a mall with a knife and the news reports a half-dozen injured ... all I can think of is that humans are much more resilient than we think, and relatively few people (even soldiers) are truly willing to slay other humans mercilessly. Perhaps at that fatal moment the blade turns, the shot goes wide, because of some natural aversion to the killing. It's hard to understand, given the way people talk about their hatreds and their determination, that they would fail to kill. But I understand it's historically been hard to get soldiers to shoot enemy soldiers accurately. Alas, I have no citation.

    4. > But I understand it's historically been hard to get soldiers to shoot enemy soldiers accurately.

      You're probably thinking of Grossman's "On Killing", a highly influential book based entirely on falsified data and misunderstanding of psychology. Robert Engen pretty handily dismantles Grossman's thesis here:


      Despite being wrong, Grossman has unfortunately, as I said, been highly influential.

    5. Interesting! Charles, thanks for the link.

  7. I don't have Basic/Expert in front of me right now, but wasn't the rule there that a half-move "fighting withdrawal" did not get a free attack, but simply turning and running did?

    Seems a reasonable compromise, perhaps.

    Then again, two combatants aren't going to be standing within reach of each other, so just turning suddenly and running full tilt would probably work.

    1. You remember correctly about the B/X fighting withdrawal. Interestingly, B/X does not include the rule on the free attack for all-out retreat. It does allow a +2 to hit and no return attack, but this seems contingent on the opponent actually chasing them. E.g.: "He decides to retreat. If the gargoyle attacks him again, his Armor Class is 3, and the gargoyle may add +2 to its 'to hit' rolls." (B25).

      As usual, we approve of Moldvay's keen consideration of the issue. :-)

    2. Also the AD&D DMG morale failure table (p. 67) has a result of "fall back, fighting", but no mechanical detail attached to that.

    3. Of course, the problem with the "fighting retreat" option is that if the enemy is pursuing, you'll never actually get way - it's only useful if you can fall back to a better defensive position, something which may or may not be available.

    4. The thing I don't like about moldvay is that it requires tracking the retreated status into the next round. That's one more thing that needs to be remembered. There is something to be said for instant adjudication.

    5. Delta said: 'Interestingly, B/X does not include the rule on the free attack for all-out retreat. It does allow a +2 to hit and no return attack, but this seems contingent on the opponent actually chasing them. E.g.: "He decides to retreat. If the gargoyle attacks him again, his Armor Class is 3, and the gargoyle may add +2 to its 'to hit' rolls."'

      I have always seen this as also depending on the initiative roll. The retreat must be announced before the initiative roll. If the retreating character (a) wins the initiative, it will move away before its foe gets to move or attack; (b) loses the initiative, its foe will attack before the retreating character gets to move away.

      I've never been certain whether this was intentional. But it is still open to interpretation. In case (a), the foe will have to move (and end within melee reach of the retreating character; but can the retreating character use its move to run away and turn at the end of the move so that it's foe doesn't get a 'rear attack' bonus after all? Or does it have to end its retreat back-to-foe so the attack is a rear attack? In case (b), does the retreating character turn its back before the foe attacks (and suffer the rear attack)? Or does losing the initiative mean that it is still facing the foe when the foe gets its opportunity to attack that round?

      Either way (naturally) the retreating character misses its attack, and the foe still gets one, provided it ends its move in melee distance.

      I lean toward the idea that winning the initiative die should be a good result not a bad one. But it's not obvious which is better. I perhaps favour the recipe 'win = retreat and turn to face; lose = cop a rear attack then run away'. But perhaps getting away is often better than ending in contact.

      I'd be curious about your reading, as it is a point about the B/X rules that I have never been clear on. That and certain interactions within the combat turn sequence.

  8. I think back attack is an unfortunate term, but I do think you are opening yourself up to some kind of attack; unless you are in some sort of formation and you are the only one withdrawing.

    I have read several times that, most of the casualties in ancient fighting came during routes which is what is being modelled. How to quantify rules that cover every situation from someone backing out of combat while his mates cover him, to all the PCs fleeing a Wight, is why the older games had a GM.

    1. Agreed that "back attack" as used here is unfortunate, and that losses in routs should be well-modeled.

      I do feel like the rout-loss is adequately simulated by a real back-attack rule; i.e., if you run, the enemy must commit to running after you, and bonuses if they can reach your back. But additional attacks at the instant of the back-out, without committing to any chase movement, seem excessive and unnecessary.

    2. OK, that did it for me. Before I was all in favor of back attacks at the point of flight. But considering them only when the attacker has committed to pursuit makes a lot more sense.

      I say this with a proviso: if the defender flees, the attacker should make the choice at that point whether to pursue or not. If he pursues, he makes his following attacks on the enemy's rear, but obviously opens himself up to focused attention from enemy fellows who may choose to pursue HIM.

      Example: A1 is in a front rank with A2, A3. They fight D1, D2, D3. There is no second rank, so there is room for anyone to flee. A1 menaces D1 and he fails morale for whatever reason. D1's action comes up, whether before or after A1's action, and he decides to flee. A1 gets no immediate attack, but is able to decide at that moment to pursue or not. If he pursues, he sticks to D1 and they're both running down the hallway.

      If A1 hadn't acted yet, he later gets his attack on D1 from behind, with bonuses. In 1st edition terms A1 is charging with an attack at the end. D1 is simply moving more than the 10' needed to close to melee, so he gets no action to attack A1.

      D2 and D3 watch them pass. They need to make the decision to pursue A1 to get a later back-attack against him. But they know if they do, they will probably find A2 and A3 charging up from behind getting back attacks against THEM. It may be a good strategy, as if they can drop A1, the three can turn and reform a rank against A2 and A3.

      If A1 charges after D1, and has already had his attack this round (he won initiative), then he won't get a freebie, but would of course get to attack D1 with rear bonuses if the flight and pursuit continued then.

      This also sort of models on a small scale what seems to happen with unit routs.

    3. I agree! I'd have no philosophical problem with that. (Maybe a bit complicated for my tastes sequence-wise, but no problem if I was a player interacting with that.)

  9. I just ran the chainmail numbers in more detail... It appears that a retreat with back to enemy situation is only likely to occur if overmatched by 7 to 1, or if failing morale when charged by horse or the Swiss pikes.

    The best rule I can make to reflect those odds is something like this:
    Fleeing: each participant in the melee rolls 1d6 per HD. For those that remain, each six indicates a back attack. For the fleeing figure, each six negates a potential attack. Horse gain back attacks on a 5-6. Figures at 6" move never score back attacks on those at 12" or greater move.

    1. That a great analysis, thank you! Gives me extra confidence that the Chainmail mass combat rules tend to be pretty well thought-out in most cases.

  10. Thanks... One minor correction. It shouldn't be overmatched by 7 to 1. It's more that rout occurs if outnumbered by 7 figures (varies by troop type, casualties, etc). Thus one figure is near certain to rout of facing 8 foes, 2 figures will rout when facing 10 (not 14), etc.

    At D&D scales, a superhero should get free swats at lone orcs... Hill giants should get back attacks on lone men at arms...

    I think my proposed rule actually still works fine if you want to mirror the chainmail odds.

  11. The only case where ground attackers can't attack slower creatures (or equal speed creatures who flee) is where the running speed is greater than the allowed charge distance. 3e for instance has only half the range on a charge compared to running speed, so a free attack is the only way to attack someone 2/3 your speed if they just keep running.

    But just let people charge as far as they can run in a round, at least once every few rounds to allow cinematic chases.

    Air combat needs simultanious strikes when the enemies happen into reach of each other, as they both just keep going after, but on the ground you stop when you attack and that works well enough. In general, if you're striding and springing or similar effects, any pass-by attack should allow a simultaneous resolution, but normal combat should be OK just taking turns if the movement limits function correctly.