Kangaroo is faster trustme, algorithm is difficult to explain at first glance.

As some other users explain before

Tame kangaroos jumps pseudo-random but deterministic from a know point like Start Base range (0x7......) or any other know value near our target publick key, so each time that a DP point (Distinguished Point with some amount of bits in ZERO in the X coordinate) is found we know its exact value.

In the other hand Wild kangaroos jumps pseudo-random but deterministic from an UN-know point like the target key, so each DP value that we found we only knows its relative value to our target key.

So the bigger dataset for tame or wild DP we have more probabilities that one item in one dataset is found in the another dataset, one that you found a match you automatically can translate all the Unknown DP into its real value this include the TARGET UNKNOWN PUBLIC KEY.

According to the birthday paradox such collision must occur soon of later so you with math/statics can calculate what are those odds for each new found value. Those odds grown exponentially.

Remember in the original birthday paradox in a room with 25 random people the probabilities that two of them have the same day of birthday are near 50%.