The Rally Point (discussion and questions)

[Swuul]

ok, is that US civil war? ACW? would they have had more sophisticated command and control and more intermediate level officers than ancient armies though? I genuinely don't know.

I don't know of how usual it was, but Romans had five mounted Tribunus angusticlavii in each legion. Their primary task was to gather up detachments of soldiers who had got lost from their main unit during combat, form from them ad hoc vexilations, and bring them back to battle. The Tribunus angusticlavii had right to perform executions if needed to fulfil their task.

To me that has always sounded like troops who had routed were then brought back to battle by the Tribunus angusticlavii. The generals had nothing to do with that in the roman army.

As for how sophisticated command and control system Romans had compared to ACW armies. The roman army had 12 ranks of officers. A legion of a bit under 5000 men had in all some 260 officers. Coincidently, an ACW Confederate division with nominal strength of 5000 men had in all some 260 officers too.

[rbodleyscott]

If the RNG generation used a normal distribution then you could allow for the most preposterous, most bizarre events occurring, while keeping the probability of them proportionately low.

Pretty much all of the RNG in the game uses bell type distributions already - i.e. two random numbers are generated, added together and averaged.

[stockwellpete]

I will note that in my second finished League game this season, once again a grand total of one non-light unit rallied from broken.

I am not sure what your purpose is in posting this comment. Other players are reporting larger number of rallies in some of their matches and we all know this happens from time to time and it can spoil those particular games. I refer you to what MikeMarchant wrote previously, "If you want to be able to have outlandish things happen, because outlandish things did happen in the real world, that's fine, but you need to ensure they happen with approximately the same regularity as in the real world." This is the crux of the issue and what we are seeking to address in the mod across a wide range of issues. Outlandish things will still happen, just not as regularly as they do now.

[Athos1660]

Other players are reporting larger number of rallies in some of their matches and we all know this happens from time to time and it can spoil those particular games.

And some other players are reporting that current rallying is just fine...

I refer you to what MikeMarchant wrote previously, "If you want to be able to have outlandish things happen, because outlandish things did happen in the real world, that's fine, but you need to ensure they happen with approximately the same regularity as in the real world." This is the crux of the issue and what we are seeking to address in the mod across a wide range of issues. Outlandish things will still happen, just not as regularly as they do now.

This reasoning is based on an unverified hypothesis : rallying happens more often in game than in real life, so it is 'outlandish', so we must fix it. Nobody can "ensure they happen with approximately the same regularity as in the real world" as we have a very limited knowledge of the past. But as you said yourself in the first part of your post, it is a just a matter of gameplay preference and perception some players have. Some like it as is. Some don't.

(edit)
However, someone with strong knowledge of History can formulate plausible hypotheses.

[MikeMarchant]

If the RNG generation used a normal distribution then you could allow for the most preposterous, most bizarre events occurring, while keeping the probability of them proportionately low.

Pretty much all of the RNG in the game uses normal distributions already.

That's interesting, Richard. Can you tell me more?

Best Wishes

Mike

[rbodleyscott]

If the RNG generation used a normal distribution then you could allow for the most preposterous, most bizarre events occurring, while keeping the probability of them proportionately low.

Pretty much all of the RNG in the game uses normal distributions already.

That's interesting, Richard. Can you tell me more?

Best Wishes

Mike

It is not quite a formal normal distribution, but it is a bell-shaped curve.

For combat calculation, this is achieved by generating two random numbers in the required range, adding them together and averaging the result.

For cohesion tests, it simply generates the sum of two 1-6 random numbers.

In both cases, it means that results closer to the average are much more common than results far from the average.

[MikeMarchant]

Pretty much all of the RNG in the game uses normal distributions already.

That's interesting, Richard. Can you tell me more?

Best Wishes

Mike

It is not quite a formal normal distribution, but it is a bell-shaped curve.

For combat calculation, this is achieved by generating two random numbers in the required range, adding them together and averaging the result.

For cohesion tests, it simply generates the sum of two 1-6 random numbers.

In both cases, it means that results closer to the average are much more common than results far from the average.

Thanks, Richard. That's good. A 2D6 roll, or similar, do produce a bell curve, but not enough of a distribution to be able to handle rare occurences reiiably. Clearly that makes the most unlikely result a 1 in 36 chance. A proper gaussian distribution would allow for 1 in a million events to occur while keeping the normal result by far the most common.

Has the 2D6 method been adopted as a consequence of this being modelled on the tabletop game? Has there been any thought about making the result calculations more sophisticated, or are there practical constraints with this?


Best Wishes

Mike

[paulmcneil]

I look forward to seeing the stats. Just anecdotally, I'm still in a game, and perhaps even about to win a game, in the DL against deserted fox because I had I think 6 rallies from broken, all at the map edge, and often of cav units, that he never had any chance of chasing down and taking care of. So, if we are going with anecdotes, it's easy to argue either way. But, like I said, I'm interested in seeing your stats!

I've just been in a game where my opponent after being out manoeuvred and largely trashed in melee got 2 Pike units, 2 hvy Cav units, and a Med Spr unit rally from rout after about 3 periods of rout each. It completely altered the game, he still lost but it absolutely knackered my winning score and made the game about twice as long as it would have been. In the same game I also got one rally from rout. So six in one game, and I'm not finding that number of rallies particularly unusual. Seems excessive to me.

what would you think about a higher chance to rally paired with only being within the general's command radius? so say 6 units rally from broken in a game, but they are all within around 4 tiles of where the battle is happening, and so quickly brought back/still close enough to be chased down by the enemy

anything that tones down the extremes within a game as far as rallies go, and is more tied to visible factors rather than the hand of God. Rallying on a Commander would be appropriate. Perhaps that's the problem, I'm an Atheist so maybe the Gods of the RNG (Blessed be its Name) don't listen to me.

[rbodleyscott]

That's interesting, Richard. Can you tell me more?

Best Wishes

Mike

It is not quite a formal normal distribution, but it is a bell-shaped curve.

For combat calculation, this is achieved by generating two random numbers in the required range, adding them together and averaging the result.

For cohesion tests, it simply generates the sum of two 1-6 random numbers.

In both cases, it means that results closer to the average are much more common than results far from the average.

Thanks, Richard. That's good. A 2D6 roll, or similar, do produce a bell curve, but not enough of a distribution to be able to handle rare occurences reiiably. Clearly that makes the most unlikely result a 1 in 36 chance. A proper gaussian distribution would allow for 1 in a million events to occur while keeping the normal result by far the most common.

Has the 2D6 method been adopted as a consequence of this being modelled on the tabletop game? Has there been any thought about making the result calculations more sophisticated, or are there practical constraints with this?


Best Wishes

Mike

We are happy with the current method. We don't want to model events that are vanishingly rare.

Of course not all RNGs in the game use a range of 1-6 for the individual random numbers - that is only for cohesion tests. Also your 1:36 assumes that there isn't another RNG "roll" governing whether the test is taken or not.

If we wanted to make the chance of rallying from rout less, we could just reduce the % chance per turn of units without a general testing to rally. This is currently 20% per turn. But of course they are testing at at least -3 modifier just for being routing, plus more if they have suffered 25% or more losses.

The chance of a largely intact Average unit (with -3 modifier) rallying in a cohesion test is 27.8%. So currently the overall chance of such a unit rallying per turn is 5.56%. Which gives it an overall chance of rallying of 27.8% before it Disperses - provided that it is not being pursued or harassed by enemy shooting.

The chance of it rallying if it has suffered 25% losses is 3.34% per turn. Overall chance 16.7% before Dispersing if not harassed.

----------------------------------

Overall, this issue is compounded by
1) Personal preference.
2) Confirmation bias. (Those who don't like the current rallying system seem somehow to perceive far more units rallying from routed than those who are happy with the current system. Unless, of course, they don't but are exaggerating for rhetorical effect).
3) Lack of tactical steps to prevent rallying, or re-rout the Fragmented enemy if it occurs. (Skirmishers. Light Horse!)

[Athos1660]

3) Lack of tactical steps to prevent rallying, or re-rout the Fragmented enemy if it occurs. (Skirmishers. Light Horse!)

I will stick a post-it with this tip on my screen.
Always saying to myself : don't let the routed enemy at ease.
Yet always keeping my light troops near the main melee.

[General Shapur]

Discussing rallied pike units. It seems to me a pike unit that flees the battle would do so after dropping said pikes. When/if they do rally, would it make more sense for them to be less well equipped. ? I expect historically they would switch to swords (which they would be less skillful with). Could that/should that be reproduced in gameplay?

[Cunningcairn]

Discussing rallied pike units. It seems to me a pike unit that flees the battle would do so after dropping said pikes. When/if they do rally, would it make more sense for them to be less well equipped. ? I expect historically they would switch to swords (which they would be less skillful with). Could that/should that be reproduced in gameplay?

Very good point. Even if you don't factor in the loss of weapons the larger the body and the more that it relies on being a cohesive unit will definitely make rallying and once again fighting effectively far more difficult than for example a skirmisher unit. I wouldn't mention it to Schweetness as he has his hands full

[paulmcneil]

Discussing rallied pike units. It seems to me a pike unit that flees the battle would do so after dropping said pikes. When/if they do rally, would it make more sense for them to be less well equipped. ? I expect historically they would switch to swords (which they would be less skillful with). Could that/should that be reproduced in gameplay?

I think this is the nub of it. Is a unit "withdrawing" or are they "routing". Routing implies chucking equipment away especially for Infantry, shields, Pikes, and other heavy equipment would all go for a Burton, whereas a unit withdrawing in good order on their own initiative is a different manner, and I can recall occasions of units historically deciding they've had enough and withdrawing from the field of engagement, but still having the ability to defend effectively against marauding Cavalry, see them off, then continue to about face and march off. So perhaps the real issue is the lack of a unit's ability to say "sod this for a game of soldiers" and march away in relatively good order, vs a rout where they just run for the hills in a panic. Such an independent state could also perhaps reflect the frequently disloyal units in Roman armies etc who would only really get stuck in if they thought they were at little risk, or perhaps stand and refuse to engage at all. The main thing I would say is to make sure all plusses or minuses in troop effectiveness are reflected in the points system.

[harveylh]

We are happy with the current method. We don't want to model events that are vanishingly rare.

Of course not all RNGs in the game use a range of 1-6 for the individual random numbers - that is only for cohesion tests. Also your 1:36 assumes that there isn't another RNG "roll" governing whether the test is taken or not.

If we wanted to make the chance of rallying from rout less, we could just reduce the % chance per turn of units without a general testing to rally. This is currently 20% per turn. But of course they are testing at at least -3 modifier just for being routing, plus more if they have suffered 25% or more losses.

The chance of a largely intact Average unit (with -3 modifier) rallying in a cohesion test is 27.8%. So currently the overall chance of such a unit rallying per turn is 5.56%. Which gives it an overall chance of rallying of 27.8% before it Disperses - provided that it is not being pursued or harassed by enemy shooting.

The chance of it rallying if it has suffered 25% losses is 3.34% per turn. Overall chance 16.7% before Dispersing if not harassed.

----------------------------------

Overall, this issue is compounded by
1) Personal preference.
2) Confirmation bias. (Those who don't like the current rallying system seem somehow to perceive far more units rallying from routed than those who are happy with the current system. Unless, of course, they don't but are exaggerating for rhetorical effect).
3) Lack of tactical steps to prevent rallying, or re-rout the Fragmented enemy if it occurs. (Skirmishers. Light Horse!)

Thanks, Richard. This is very helpful and makes me feel better about the current rally system. Does shooting at a routed unit keep it from rallying that turn?

Harvey

[Morbio]

Can you provide an example of troops rallying out of sight of the main battlefield and therefore causing their entire army to rally? Any battle between 3000 BC and 1500 AD will do.

I know of two offhand.

Battle of Magnesia (Romans vs Seleucids). Livy describes Seleucid cavalry breaking the Roman left which proceeded to stream back to their fortified camp. A tribune in charge of the camp went out and supposedly rallied the fleeing mass by initiating executions of the routers by the camp guards. Apparently it was successful and with Macedonia and Thracian troops prodding them back, they re-engaged and checked the Seleucid cavalry threat.

Battle of Verneuil (Lancastrian English vs French). Accounts a fuzzy as always but apparently the Milanese cavalry initiated a massive charge on the English right-wing which was made up of archers. The dry and hard-packed soil prevented their now standard defensive earthworks and stakes and the Longbowmen were promptly overrun. Many fled and were later executed for treason. The Milanese cavalry proceeded to run off the battlefield to try and loot the English baggage trains and with the threat gone, the English right flank regained their composure and at some point in the battle rejoined part of the main battle and attacked the French infantry which was locked in a stalemate with the one contingent of English men-at-arms (the main battle line had split in two at this point). The French apparently tried to flee shortly afterwards and were either drowned in the nearby moat or massacred. The English then went back to surround the Scots which were still facing off against a separate division of the English army. Apparently no quarter was given to the Scots.

Thanks for answering Mike. The Magnesia example isn't exactly what I was referring to as there was "friendly interference" in the rally. However that aside, it still only gives 2 examples over a period of approximately 1800 years. And that is the point. The probability of it happening was exceptionally low.

While I tend to agree, I think this is a bold statement just because Mike only listed 2 examples. I don't think Mike really was stating this was all, it was just 2 offhand, and I don't suppose all the examples in history are either documented (with accuracy) or that anyone would know of them all if they were documented.

Going back to one of the earlier comments about double/triple drops. Again I agree that it probably shouldn't happen from a fallback, but who is to say that this never happened in history? Presumably there was some threat on the lancers so falling back would have some risk, which could cause troops to panic and thus lose cohesion and whenever a panic happens it could spiral out of control. There are many examples in real-life, even if not military, where a small issue leads to panic and then catastrophe.

Many of the contributors to the forum don't want predictable chess-like results, so are happy to have RNG determine results. If you have RNG then you will get edge cases that can be catastrophic... and who is to say that this wasn't realistic in many battles as a detailed report of all the events doesn't exist in anything except 20th century onward and even then it is probably sketchy at best and may have bias from the reporter.

It makes me smile that so many people want a realistic game... based on what they think is realistic and there is no solid basis for what is realistic!

[Schweetness101]

Discussing rallied pike units. It seems to me a pike unit that flees the battle would do so after dropping said pikes. When/if they do rally, would it make more sense for them to be less well equipped. ? I expect historically they would switch to swords (which they would be less skillful with). Could that/should that be reproduced in gameplay?

I like this idea. I'd considered looking at how difficult it would be to swap out a unit's unit type mid battle (ie in this case from pikes to say swordsmen after rallying from routed). This would also be needed for dismounting units mid battle. But, given that you cannot dismount units mid battle (and thereby change their unit type), and that being able to do so seems like it would be desirable, I assume the technical implementation was considered to be too difficult to be worth it. Although maybe not and that was just a design choice.

[stockwellpete]

The chance of a largely intact Average unit (with -3 modifier) rallying in a cohesion test is 27.8%. So currently the overall chance of such a unit rallying per turn is 5.56%. Which gives it an overall chance of rallying of 27.8% before it Disperses - provided that it is not being pursued or harassed by enemy shooting.

The chance of it rallying if it has suffered 25% losses is 3.34% per turn. Overall chance 16.7% before Dispersing if not harassed.

These seem quite high chances to rally to me. 1 in 4 chance if the unit is at 75%+ strength and 1 in 6 if it has been badly mauled. Perhaps the differential between the two states could be wider? 1 in 4 for units that have not been too badly damaged and 1 in 10 for those that have?

Overall, this issue is compounded by
…..
2) Confirmation bias. (Those who don't like the current rallying system seem somehow to perceive far more units rallying from routed than those who are happy with the current system. Unless, of course, they don't but are exaggerating for rhetorical effect).

So we are being dishonest again, are we? We are not exaggerating. We are just saying it is a factor in some of our matches. I am not actually sure how often it happens, but it happens enough to get noticed.

[Schweetness101]

another thing to be clear about is that the alt gameplay mod, if that is even what is being discussed, would not necessarily be aimed directly at decreasing the net total rallies in any given game. In fact, it could end up with more rallies potentially because I have increased the chance to take a test to rally for a broken unit on any given turn (if it is not being pursued), BUT those rallies would only happen on units with a general, or within the command radius of their general.

So, what rallies from broken do occur are more likely to be near the main battle still, or the result of active player intervention to send generals back to rally units intentionally. I think this could retain the concept of rallies and their frequency, while reducing the frustrating and seemingly unrealistic aspect of units on the far off map edge rallying to fragmented, and staying that way far away from the battle where it doesn't really make sense that physically or psychologically they would still be contributing to the not-routed percent of the army.

Just trying to clear up here that I don't think anyone is advocating for getting rid of rallies. They are one of the more compelling and historically accurate parts of FOG that make it fun and a nice simulation. I just think it is interesting to experiment with alterations on how they are done.

[MikeMarchant]

Of course not all RNGs in the game use a range of 1-6 for the individual random numbers - that is only for cohesion tests. Also your 1:36 assumes that there isn't another RNG "roll" governing whether the test is taken or not.

Yes, of course, but that doesn't alter the fact that once a cohesion test has been called for 1:36 is the limit of the least likely event.

I think my argument is probably misplaced, because it's not really the probability of an event occurring in isolation that bugs me, and I expect that might be true of others too, but the clustering of such events. If an event is very unlikely, then clearly it will still occur (hopefully rarely), but when there is a string of unlikely events, either occurring simultaneously (or nearly so) or over time but associated with an individual unit, that's what causes distress.

That's probably not very clear, so I'll try to illustrate the point.

Events over a period time, occurring to the same unit might include the situation where the unit which has a 58% chance of winning the melee fails to do so 18 melees in a row. The chance of that happening is a little less than 1 in 6 million (if I recall millennia ago maths lessons).

Event occurring simultaneously might be the situation I suffered recently, where a tight game, late on, with a small advantage to me, collapsed in a single turn because my opponent rallied almost all of his units eligible to rally. Not only those routing, but also those disrupted and fragmented, and then to add salt to the wound, he also won every single impact and melee on his turn, even those where the probability of that was highly unlikely. A tight game turned into a disastrous loss as a result of outlandish luck. Has that ever happened on a real battlefield? I’ll have to leave that question to the historians, but it certainly never happened in any of the limited number of battles I am aware of.

There often seems to be a disparity between the probabilities reported by the game and the results seen on the field of battle, but I say seems for good reason. I am beginning to suspect that the fault may lie not in the game mechanics (although we can still argue over the likelihood of any particular event occurring), so much as the fallibility of computers' random number generators hitting irregular patches.

I don’t know how this weakness can be overcome, unless there’s a way the game could create some novel technique to create a genuinely random number. I expect if that was possible whoever was responsible would become very rich. Beyond my ability, sadly.

So perhaps it’s not me who suffers ill fortune in war, perhaps it’s my computer. I’ve read many computer reviews, detailing many technical specifications, unfortunately luck has never been included amongst them.


Best Wishes

Mike

[rbodleyscott]

There often seems to be a disparity between the probabilities reported by the game and the results seen on the field of battle, but I say seems for good reason. I am beginning to suspect that the fault may lie not in the game mechanics (although we can still argue over the likelihood of any particular event occurring), so much as the fallibility of computers' random number generators hitting irregular patches.

I don’t know how this weakness can be overcome, unless there’s a way the game could create some novel technique to create a genuinely random number. I expect if that was possible whoever was responsible would become very rich. Beyond my ability, sadly.

It is very unlikely that there is any such weakness in the random number generator algorithm.

The phenomenon you are seeing is called Poisson clumping, and is a feature of all random events, not just computer generated ones.

https://en.wikipedia.org/wiki/Poisson_c ... r%20bursts.