Additionally, it's important to know how to confirm on a play a team member has made. It's easy to see a Lux binding land and want to go all-in. However, sometimes that's not a good play, especially if the enemy has counter-initiation available.
The obviously lost battles
You can tell your team has lost a fight in a lot of ways, and often you can tell before any real damage has gone down. Once you have some experience in how various champion abilities impact a fight, seeing them used will tell you a lot about how a fight will go down.
A simplistic example is if, for instance, the enemy Annie has her stun available, fires Tibbers erroneously, and ends up having little impact (either from missing, getting spell shielded, Cleansed, etc.). It is very likely that the opposing team will win a subsequent engagement if it occurs shortly afterward. Annie will need to charge up her stun again and wait on the cooldown for Incinerate to have any hope of salvaging the fight for her team.
A "perfect" initiation, by comparison, pretty much always means a won fight. If Amumu lands Bandage Toss on your ADC (do I really have to call them "marksmen" now?) and lands his ult on two other people, you've probably lost that fight. Even if you manage to fire a counter-initiation, you would need a pretty big wombo combo (some Orianna/Sona combo ult or something) in order to stop that. And if you did have something like that, the enemy can just DPS the people who got caught and not fully engage, and all you can do is blast Amumu and your ADC can't shoot at all (if she's even alive).
Those are fairly obvious scenarios, but the takeaway is that you need to look at what tools have been used and what tools are left on the table. In the Annie example, it might not matter too much if your team still has an Orianna ult available, especially if the enemy support doesn't have a teamfight ultimate (maybe it's Soraka or something). In that case though it's more about what your enemies don't have; if they never had a good initiation tool to begin with, then losing yours makes the odds more questionable but not obviously lost.
In that case, my team lacked a solid teamfighting ultimate, but we did have a lot of single-target damage and picking power, plus Singed being nigh indestructible. I used those strengths to make a play that would win the fight after the enemy had spent its major trump card. If I had just gone in without being patient and waiting for Veigar to screw up, we would have lost the fight.
On the flip side, it's important to cut your losses if something bad does happen. Let's say that you miss that critical ultimate. It's important that your team not fight there! If you throw Crescendo down and hit absolutely nothing, your team will probably lose the next fight, so you are better off conceding map control for the next 90 seconds and not giving up a bunch of gold and experience for fighting and losing.
Of course, there are exceptions. If the enemy team is posturing for Baron and you blow a teamfighting ultimate, your team needs to shift gears and try to do its best to keep the enemy team from killing Baron. You may not be able to win a fight in that situation, but you still have a chance if you fight while they're fighting Baron, so you may be able to do something. You should still avoid fighting directly unless an advantage presents itself, though.
A fight has started; blood is getting spilled. In most cases, the team that wins the fight in the end is usually the team whose damage dealers live the longest. There are some cases when it comes down to tank versus tank (usually hilariously), and you may have to learn whether you can win a fight in that case. There are also some really skewed cases such as ADC versus Rammus, but generally the team with living damage dealers wins.
This doesn't mean "dive the DPS" because that will get you killed. What it means is that if your DPS dies early in a fight, you should strongly consider disengaging unless your team has an overwhelming health lead. Anyone who stays and fights or cannot disengage is probably already dead.
Here's another extreme example: If the enemy Katarina is allowed a solid initiation and is able to land a clean ult, she will probably chain-kill everyone low on health unless folks drop out of the fight immediately or Katarina is slain immediately afterward. A similar situation occurs with Kha'zix, whose evolved leap resets on kills or assists and can chain-murder your team. It's a very good idea to disengage from that lost fight or you will feed Kat or Khaz or Trist or Yi another reset and more chances to keep going.
A more reasonable example is when an ally is slowed and running but has no real hope of escape. Sometimes you can help. But too often, helping will get you killed too. It's critically important that you understand the difference. If you can grant your ally bonus movement speed, cleanse his debuffs (Mikael's Crucible or some similar thing), or stun/knock the enemy, you might be able to help. However, I've seen Ezreal turn around to help an ally and get picked off first, followed by the person he was trying to help. If he had escaped, he might have been able to use his ultimate to clear a push wave or attempt to steal a map objective. In this case, Ezreal lost my team a tower and fed shutdown gold to the enemy.
The moral of the story is to not give the enemy more than your team has already given away. Sometimes this does mean engaging, though. If you stand to lose an inhibitor turret or Baron if you don't stand and fight, you might need to fight. But if you can cripple the enemy team members enough that they can't push or do Baron, it might be the best way to cut your losses.
We understand what it's like to climb the skill ladder in League of Legends. The Summoner's Guidebook teaches you the tools you need to get a competitive edge. Whether you're climbing the ranked ladder, playing Draft Dominion, or getting crushed by intermediate bots, every enemy has a weakness. And every Thursday, Patrick Mackey shows how you can improve improve on yours.