# Why does the tip-selection prefer new tips?

Most of the issues regarding slow or never confirmed transactions seem to arise from time-intensive POW and a badly connected full-node, which both lead to an attach-position not at the tangles "front".

So why does the tip selection algorithm favor new transaction? What are the reasons and mechanisms that led to this decision? Why isn't it better to confirm the old ones first?

It must be obvious for everyone that attaching a new transaction to a transaction that was already validated is useless for the network.

It must be obvious for everyone that a node should do it's best to increase overall TPS (transaction per second).

When selecting tips, a node is not sure that the tip is still a tip. Due to network lantency, there is a probability the what the node see as a tip is already validated in an other node.

This probability to not being a tip anymore increase with time. i.e. there is high probability that an old tip is already validated in another part of the tangle. That's why a node will prefer very new tips over old ones.

(source : whitepaper page 7-8-9)

• Might be worth shedding some light not only on what the "probability of not being a tip anymore" is, but also on why the probability is what it is, as this would cater to the actual question of the OP of "why isn't it better to confirm old ones first". Dec 15, 2017 at 0:36
• Not sure to understand your comments. A node don't confirm old one first because there is high probability that another node already did. Dec 15, 2017 at 0:48
• But isn't that a problem, because (when driven to an extreme) one would only ever reference the very last (known) tip, and thus leave "older" ones behind. And as a consequence the tangle would continuously accumulate tips. Wouldn't it be better to fan out and approve tips from a wide range? Dec 15, 2017 at 0:56
• In addition to @Phil-ZXX, the full node should know every transaction of the tangle after a certain time. Therefore, couldn't it find the remaining tips and select a certain percentage of them? Dec 15, 2017 at 9:54
• @Phil-ZXX "when driven to an extrem" is IMO not a valid assumption. The tip selection is driven by probability and won't go to an extreme (unless the node handle almost zero transactions and in this case it is not a problem at all) Dec 17, 2017 at 10:13