When playing a RTS (skirmish - single player) usually there is a moment after which the player holds most of the map, most of the resources and his units compared to the AI are 7:1 or something like that. In this situation the possibility to loose is very small, almost impossible. But the enemy still has power and the player must destroy it playing 10, 20 or more minutes.

Scenarios:
1. Win when you hold 70 or more % of the resources (or territory) - as in Rise of nations. That is not valid for the AI, because the player could survive with only one base... maybe.
2. Let the player smash the enemy, the situation will become more and more hopeless for the AI. Maybe that will be fun or maybe not?
3. The enemy will have hidden feature when is in bad situation to pull out very special power/unit so it can crash your army and then the gameplay will become more dynamic because it will take a lot of territory and a lot of players unit will be destroyed, the player can even lose. But what the next time - the player will get the territory around the enemy base again, should this happen again? And maybe the same for the player - when he is in bad situation holding only his base to receive bonus?
In the chess at the end of the game, endschpill (?) figures are less and usually most units are destroyed, so even the King could do something. But in RTS usually the winning side will have a lot of powerful and upgraded units. And not even at the end, usually there is a point after which A will become stronger than B and that can not be changed easy.

If I'd have a 7:1 advantage, my thought would be that I did not attack early enough (forgot or intentionally toyed around). Prolonging the already derailed game even further in that case IMHO won't be fun, same as depriving the player of being able to overrun the opponent.

2 hours ago, Pleistorm said:

usually there is a moment after which the player holds most of the map, most of the resources and his units compared to the AI are 7:1 or something like that

Ideally you want the game to end very quickly after one player has an insurmountable advantage. Otherwise the player gets bored mopping up (that feeling in StarCraft when a player hides their command centre at the edge of the map, and refuses to concede...).

You can have a upkeep mechanic built into the game, something like ammo, food or morale. Once a player is backed up into a corner they either starve to that mechanic or make a last ditch effort to gain more resources. Either way the game ends or an interesting final conflict happens.

For games against the AI, as swiftcoder said you should have the game end fairly early into the "insurmountable advantage" part. Or "end" at least as in the game informs the player that they have won, though they can then elect to continue their game and do the mop-up part if they're that sort of player. You can mark this game-winning point in more interesting ways than simply a "military score" or some such. For instance, once you have reached this point, perhaps every AI faction left will band together and throw everything they have at the player in a last-ditch effort to regain some ground (or use some mega-weapon as you mentioned). Assuming the player wins this final battle without much damage, they are given the win. Basically a final challenge to the player to make the end come with more of a bang than a whimper.

Depending on the game type and nature, you could have other manner of end-game challenges. Rebellions amongst the players own troops creating a sudden new enemy faction that is less underpowered, perhaps.

One tactic in games where selling buildings is allowed, is for the AI to "resign" by selling all their buildings (which usually converts them into a small group of military units) and sending all those units at their opponent all at once.

It gets rid of the tedium of mop-up, and also provides for a final opportunity to overwhelm their opponent with a single mad rush.

Thanks for the replays.
I was usually playing Cossacks and Company of Heroes to the end but recently I am playing my game only and when I smash the enemy to the corner it is some kind boring to keep on, maybe because it is my game. (But usually the enemy is smashing me.) Finishing the game is a hard task because enemy tries to break the supplies and that way the AI can destroy a big army but it doesnt happen after that win-point of time.

The idea of rebelions could work but actually the player is using machines so they can be reprogrammed by the enemy at any stage, will try that as mass effect at the results.
Selling buildings is impossible due to the character of the game.
I put a mechanism when the player has a lot of units the income will be reduced but that lead to my defeat, it is hard to do because the progression of the enemy. During the gameplay the enemy is causing a great damage to my troops and reviving is hard. I think the main reason is that I dont have upgrade tree as in other games but upgrades are locked to the units and buildings and when loosing important things or a lot of things it is hard to keep it up.

Probably the best solution is to give a chance to the player to choose when it gains 70% of the territory to win the game (and if he wants can keep fighting as Telcontar wrote) or to receive a challenge (as the big Boss in arcade games), if he lose the battle the game will keep going (if the damage caused by the Boss is not huge) and eventually if this happens again, again some big Boss; or not any challenge... And maybe an option when the player is in the bad situation to receive a Boss or pack of resources...

I've only seen early ways of winning in turn-based empire-builders, where the a colony(/city/province) has extensive defensive bonuses when developed and can easily maintain(though not produce) plenty of defenders, in RTS defenses are non-moving units who generally are NOT stronger than moving units, their advantage is solely that they do not require upkeep.

How about a mechanic that accelerates the "spiral of death"?

For example, in the game Z: Steel Soldiers, the map is divided into territories.  The more territories a player controls, the faster they're able to build and deploy new units.

In this way, an advantage rapidly becomes overwhelming. Once you control more territory than the opponent your increased production rate makes it increasingly easier to defend your existing territories and capture more, which in turn gives you even more of an advantage.

Of course, this exact mechanic may not suit every game, but an accelerating spiral of death is something worth considering if your end game is feeling too drawn out.  The board game Monopoly employes a similar mechanic with property ownership and rent.

In a (theoretically) ideal game, you want both comeback mechanics in the early/mid game, and an unstable equilibrium in the end game.

Why? Because without the comeback mechanics you'll have frustrated players who made one or two mistakes in the early game, and have no chance of catching back up. And without the unstable equilibrium, a player who is clearly losing by a small margin may prolong the game almost indefinitely.

Sometimes you apply these via direct rubber-banding (i.e. a MOBA that gives bonus XP when a team which is losing manages to achieve a kill). Or by encouraging high risk/reward tactics for players who have an advantage. Or build it into the tech tree, by making early units have distinct tactical tradeoffs, and late units somewhat overpowered...