Combo sideboards are the hardest to build, because taking a card out of your deck is disruptive to your game plan—every piece is there for a reason. It’s like the aggro decks, except multiplied by a lot. There are many matchups in which you do not sideboard at all with a combo deck. For this reason, combo decks usually have great sideboard cards against other combo decks—they can play 3 of each very specific hate card because they can never bring in more than 3 cards in any matchup anyway.
There are 2 possible approaches when it comes to combo deck sideboards: Either you sideboard to beat their board, or you transform. Sideboarding to beat their board is straightforward—it’s saying “I know you have hate, but I’m going to board something that beats your hate.” Let’s take a look at Ross Meriam’s Dredge deck:
This sideboard has 6 cards that deal with hate pieces—Claim, Decay, and Grudge. The goal here is to create a spot where, if they don’t draw their hate, you win. If you both draw your sideboard, you also win, and then if they draw theirs and you don’t draw yours, you lose. You can’t have many answers because you’re diluting your deck, but having about 4 of them means you think your deck is better enough that. If you don’t lose to the hate cards, you’re going to win even with 4 dead cards in your deck.
The second approach, the transformative sideboard, is even tougher to pull off. It consists of completely changing your game plan in a way so that your opponent’s sideboard is not very effective. It used to be more common in high-level events, such as when Kai Budde played 3 Morphlings in his Illusions/Donate deck and morphed into a control deck post-board when people brought in enchantment answers. It still exists nowadays, and when it works it’s a very powerful strategy because it creates a discrepancy in knowledge that is not seen anywhere else. One example is Vincent Lemoine’s Storm deck from GP Lille:
This is deck relies on an enchantment and on the graveyard to win. When Vincent expects too much disruption for those elements, instead of trying to fight them (with something like Negate), he changes his plan altogether. He boards in Thing in the Ice and/or Empty the Warrens. Those cards are synergistic with what his deck is already doing, and offer him a different angle of attack. If you play against Vincent, you never know how to sideboard, because you don’t know what he did. You might take out all your Path to Exiles and lose to Thing in the Ice, you might take out your Anger of the Gods and lose to Empty the Warrens, or you might keep both of those in your deck and still lose to Past in Flames because you have so many dead cards and he didn’t board anything.
The cards are different, but the idea is the same—if you think they’re going to target you with graveyard hate or spell-hate (such as with Eidolon or Ethersworn Canonist), you can morph into a separate combo deck that is not affected by those cards.
When you’re playing a transformational sideboard, it’s very important to hide how many cards you’re siding in and out. If they know you only sided in 2 or 3 cards, then they know you didn’t morph—if they see you boarding in 10 cards, they know you likely did. Even more importantly, if they see you change nothing between games 2 and 3, they know your deck remained the same. With this type of deck, you can’t be lazy—you have to at least pretend to sideboard between every game. The best way to do this is to usually shuffle your sideboard in your deck and pull out 15 cards, but you can try to do it normally while just being sneaky about it.
One thing you have to be careful with transformative sideboards is to not depend on the same resource that they’re trying to attack. If you have a Dredge deck, then it doesn’t make sense to have a board plan of Tasigurs and Tarmogoyfs because those cards will still lose to any Dredge hate they might have. It’s the same thing as playing Affinity and having Torpor Orb, something that I never found to be particularly effective because they’d have the answers against you anyway.
This is of course very generic, but the goal is to help you when you don’t have a ton of time to test every possible combination. If you follow the guidelines here, then at least that will narrow down your testing process a little bit.