
Some math is needed to see what this card actually does. Please forgive any errors-- I did my best!
-
Preventing the auto-fail versus causing it. Drawing the auto-fail and no other eligible symbol tokens would be a bad result for this card, since you have to choose auto-fail in this case. So what are the chances of this? I calculated that it is 4.8% with standard NotZ chaos token pool. This is based on the chance of getting the autofail in 5 token draws (31.25%) multiplied by the chance of drawing none of the other indicated symbol tokens with the remaining 4 draws (15.4%). Compared to the chance of drawing an auto-fail with 1 token draw (6.25%). this is a 1.4% absolute risk reduction, which means you would have to play the card on average 71 times before it actually prevents an auto-fail result. This is a little hard to wrap your head around, but easier to understand by remembering that the chance of drawing the auto-fail is already low, and this card does have some risk of forcing the auto-fail. In summary, I don't think this card can be considered a way to prevent the auto-fail token.
-
Increasing the chance of getting a symbol to activate a player card effect. This includes the following cards (I'm ignoring a few like Ritual Candles): Hypnotic Gaze. If there are 5/16 symbols, this increases the chance from 31.25% to 89.4%, an increase of 58.2%, which means that almost 3 out of 5 times that you play this, it will give you a counterattack that you would not otherwise have gotten. This might be worth it in certain decks (possibly Diana Stanley, whose ability partially offsets the cost). Jewel of Aureolus. Odds are comparable to Hypnotic Gaze, this giving you the +1 card or +2 resources an extra 2 out of 5 times. But this one can be combined with other effects (including Hypnotic Gaze). So if you are already using the Jewel, Dark Prophecy becomes more attractive. Counterspell. This one is a little different, since you don't have to play the card unless you need it. But 2 out of 5 times you play Dark Prophecy you will generate a suitable target for Counterspell over the background odds, in which case you can decide to use the Counterspell. There are 3 outcomes from this: you get get no symbols so you can choose a nice token and save your Counterspell, you force an auto-fail (see above), or you get a symbol token and decide whether to counter it. I think if you are already using Counterspell, this makes Dark Prophecy more attractive. Defiance: If there are 2 skull tokens (or 2 of another symbol of your choice) out of 16 total, Dark Prophecy increases the chance of drawing one from 12.5% to 54.2%, a 41.7% increase which means 2 out of 5 times will create a new hit for you. But the "hit" is basically negating the chaos token, so I don't think this is worth two cards. Song of the Dead: Like Defiance, the Prophecy gives a 41.7% increase which means 2 out of 5 times will create a new hit for you. But this is just not worth it, considering Shrivelling and Shards of the Void are available, which will do as much average damage even if you had a Dark Prophecy for every attack. Jim's Trumpet. Same odds as the previous two, meaning you add a horror heal 2 out of 5 times more than the background odds. I think you'd have to be desperate to think this is worth it. (See below for general use in Jim Culver decks.) Recall the Future: Since you should choose a symbol, often the skull, this has similar math to the previous 3 above, thus granting you +2 skill an extra 2 out of 5 times. Not worth it, especially since the Recall the Future is then exhausted and can't be used for the rest of the turn!
-
Avoiding tokens that are worse than the indicated symbols. This is presumably the main use of this card. Let's use the standard NotZ chaos token pool, and using The Gathering symbol definitions with 2 or less ghouls in play, in which the -3, -4 and auto-fail tokens are worse than the negative symbols. The chance of getting one of these in 1 draw is 18.75%, while the chance of getting at least one token that is not one of these is 100% (since there are only 3 in this example). So this reduces the chance by 18.75%, meaning that one out of ~5 times it would prevent a negative result. But one should keep in mind the fact that you have to select a negative symbol, which could be only an auto-fail (4.8%, see above). This effect is more useful when there are lots of really harsh numbers (like on expert mode) but when the effects of the symbols are relatively mild. Often the symbols are at least as bad as the negative number tokens, which would make this effect less useful. The exception is Jim Culver, whose modification of the skull token swings this in your favor. Otherwise, I'm not impressed with this.
My conclusion based on all of this is that this card is mainly useful with Jim Culver and perhaps occasionally in a deck that uses cards with good synergy (Hypnotic Gaze, Jewel of Aureolus and Counterspell, etc). This card should not be used in most decks and I don't think you should base any deck around it.