This is only one of the issues. People are making assumptions and models with no relation to the system.

They could make a hypothesis of the system and test that, but where is the sensation in it?

elliotolds [5:35 AM]
csw's description does not make sense to me. csw says "Gamma is set and modeled using using a low kappa Power Distribution network. Bitcoin is an extremely high kappa Poisson distributed network."   The selfish mining paper doesn't model the network topology. Gamma is just the fraction of honest miners who mine on the selfish miner's block. csw seems to think that in practice gamma will be low, but the selfish mining paper still shows selfish mining is profitable when gamma is 0.

csw [5:59 AM]
And that is also wrong, but I am doing one point at a time.

[5:59]
Seeing as you keep jumping, it just means nothing is sorted.

[6:00]
Start looking at each area.

[6:00]
Lets work on the first assumption. Then we will work on what the inconsistencies with the rest of the model.

[6:01]
At no point have I said, this is it - now it is all wrong. I am TRYING to step people through an extremely complex area and have them learn step by step.

[6:03]
And again, the model is wrong, but until you start to actially understand the nature of the network, then it is no good starting to debate the maths.

If you have the wrong model, it does not matter how good the maths is... And it is also flawed, but one thing at a time.

[6:03]
IFF Bitcoin was a low kappa power law system, then they would have some argument. It is not.

csw [6:17 AM]
@elliotolds
What is the effect of changing from a Poisson to a Power law network?
What is the distance and how does this change centrality measurements?
In Babaioff (2012) what was the study of Gamma referring to?
How were these findings used in the Selfish Miner paper?
If Bitcoin was degraded to a Small World system, what would change?

Babaioff, M., Dobzinski, S., Oren, S., Zohar, A.: On Bitcoin and red balloons. In: ACM Conference
on Electronic Commerce. pp. 56–73 (2012)

[6:18]
If you cannot answer those questions, you cannot even start to answer the assumptions used in SM.

[6:23]
Let's think on Percolation theory...

It the propagation of blocks and Transactions *Subcritical* or *supercritical* or something in between?

[6:23]
What is the Sharp threshold for Bitcoin's Percolation model?

[6:24]
A little more reading:

Smirnov, Stanislav (2001). "Critical percolation in the plane: conformal invariance, Cardy's formula, scaling limits". Comptes Rendus de l'Académie des Sciences - Series I - Mathematics. 333 (3): 239–244.

[6:28]
Next, to go back to why this matters...

The first transaction released is on an open network with large inter-connectivity.  It is an SEIR-C model and SIR as a simplification. The nodes that have received a block are not.

Once you have recieved a TX, that node is now immunised from a competing block.

Node response:
 1 Receive a valid TX or Block
 2 Reject an equal but competing block that comes even a second later

As the nodes reject the new block, they do not forward it

This means that the distance and centrality values for the first node to release are NOT the same as the later ones.

You cannot model the transmission using the same formula and data... Bad assumptions lead to bad models. It is simpler, but wrong.

csw [6:32 AM]
uploaded this image: image.png
Add Comment

csw [6:34 AM]
As the SM reacts, they are on the "pruned" network graph. The nodes the SM can send to are the ones who have not received the initial release of the TX or Block.

[6:35]
So, you cannot use the same probability for each. Conditional probability needs to incorporate the Bayesian prior.

mwilcox [7:12 AM]
Hmm a bit to catch up on in here

csw [7:12 AM]
Oh and it seems to work, but in the same way that a Martingale strategy does,
https://www.forebet.com/en/betting-systems/88-the-martingale-fallacy

https://en.wikipedia.org/wiki/Martingale_(betting_system)
forebet.com
The Martingale fallacy
Considered by many a good and successful betting strategy, the Martingale has put a lot of people into a rather uncomfortable position. The forebet team believes that the Martingale strategy is misunderstood and it is not advisable to use it. If you follow this strategy you will have to double your bet after every loss so that the first win would recover all previous losses plus win a profit equal to the original stake. After the first winning bet you continue with bets equal to the last one. We Show more… (49kB)
 Wikipedia
Martingale (betting system)
A martingale is any of a class of betting strategies that originated from and were popular in 18th century France. The simplest of these strategies was designed for a game in which the gambler wins his stake if a coin comes up heads and loses it if the coin comes up tails. The strategy had the gambler double his bet after every loss, so that the first win would recover all previous losses plus win a profit equal to the original stake. The martingale strategy has been applied to roulette as well, Show more…

[7:12]
SM is a poor Martingale...

mwilcox [7:13 AM]
> What is the condition for "unbounded spread" in Bitcoin?
That nodes have an economic incentive to both receive and to broadcast transaction data, that there is no limit to the number of nodes that can join the network
> Is this directed or undirected?
Undirected. Each node can send or receive transactions, forward or receive blocks, replace or remove, and can both validate or invalidate blocks by choosing which head they mine on

csw [7:13 AM]
SM is known as a limited response system anti-martingale.

[7:14]
Now, you are correct and also NOT>

Bitcoin is undirected.
The Selfish Miner is directed.

mwilcox [7:15 AM]
Yes

csw [7:15 AM]
Think why and if you need to, re read the paper.

[7:15]
:slightly_smiling_face:

mwilcox [7:15 AM]
Because they do not forward tx.

csw [7:15 AM]
So, you cannot use the same maths and assumptions there either. The authors have modeled the SM response as the same undirected network....

[7:15]
:slightly_smiling_face:

mwilcox [7:16 AM]
Got it

csw [7:16 AM]
And they hold and also block blocks

[7:16]
As I said, you need to have the assumptions right or a model is just a pretty example that you can do undergrad maths

mwilcox [7:17 AM]
the SM is a giant component? they stay constant because they always publish empties?

[7:17]
or all SMs and their nodes?

csw [7:17 AM]
The SM network is an overlay power law distribution

[7:17]
The SM would need to have at least 50 % to be the giant network

mwilcox [7:17 AM]
right that makes sense

[7:17]
hmm

csw [7:18 AM]
So, we are again back to, if you get more than 50% you can attack the system

[7:18]
Not 33%


[7:18]
I can see why people do not understand  this, and I have benn a little lax. I had too many other things to complete.

[7:19]
But I am back to teaching how all this works and I am TRYING to move past analogy...

[7:19]
But I am terrible at just answering as the answers require that you have knowledge of the system

[7:20]
This is also an issue with SegWit, They have moved to a system that changes the network incentives and it changes the connection graph

[7:21]
Bitcoin is very carefully strung together. It is to as easy as people think to just play with the system and have it work long term.


mwilcox [7:22 AM]
Yes people think of 'Bitcoin' as just some software that they can update / add features to, as opposed to a set of continuously applied rules with very carefully balanced game theory at the core of its design


mwilcox [7:41 AM]
Ok so I'm just reading the SM paper again (been a while..)

[7:41]
things that jump out to me

[7:42]
>"incentivizing rational miners to join the selfish mining pool"
(edited)

mwilcox [7:42 AM]
does that not make the 'selfish pool' simply share their blocks with those who join
1 reply Today at 7:46 AM View thread

mwilcox [7:43 AM]
>"In this experiment, we use the simulator to simulate 1000 miners mining at identical rates"
>"We assume block propagation time is negligible"
not the best way to simulate a heterogeneous network (edited)

[7:44]
>"In the first scenario where the honest nodes succeed in finding a block on the
>public branch, nullifying the selfish pool’s lead, the pool immediately publishes
>its private branch (of length 1). This yields a toss-up where either branch may
>win."
Er, how could it be a toss-up, when the selfish-mining pool already received the block

csw [7:46 AM]
_"does that not make the 'selfish pool' simply share their blocks with those who join"_
Yes.

It is a secret strategy that is told to one and all.

One person can have a secret.... Two, three...

[7:47]
Starting to see the flaws and the incorrect assumptions? (edited)

[7:51]
_"In the first scenario where the honest nodes succeed in finding a block on the
public branch, nullifying the selfish pool’s lead, the pool immediately publishes
its private branch (of length 1). This yields a toss-up where either branch may
win."_

But,it s not a toss up. It is not instant. The Selfish miner needs to wait and validate a block from the Honest Miners. The Selfish miner has a response that is conditional to seeing the block.

And they assume the same node connectivity, not seening that the release of the block from the Honest nodes starts to innoculate the system against a second block (no time stamps).

[7:51]
They also need to ensure that no pool member defects.

[7:52]
Remember, a pool member finding a block needs to be paid.
So, they are "in" on this or they are not and want money now...

mwilcox [7:52 AM]
exactly

[7:52]
also what if all pools are selfish mining ?

[7:52]
it just self balances

csw [7:52 AM]
This is a further cost.

[7:52]
Now, if ALL pools selfish mine, then it ends as a cluster fuck

[7:53]
But the model is different again

[7:53]
The economics of all miners doing this are the same if they do not. That is, it is no more profitable than not selfish mining

[7:54]
But, a pool member can start to earn by defecting...

[7:54]
So, a dishonest selfish miner can make more profit collapsing the system...

[7:54]
And the SM loses money...

[7:54]
Which leaves them to stop Selfish mining :slightly_smiling_face:

mwilcox [7:54 AM]
it's just batching. its too risky since if you aren't getting a proper share for your hashpower (because the pool owner messed up its publication pattern) then you're gonna go to the place with the more reliable hashpower

csw [7:54 AM]
:slightly_smiling_face:

mwilcox [7:55 AM]
if a pool owner is selfish mining, they'd take any extra profit for themselves anyway

csw [7:55 AM]
Starting to see it now :slightly_smiling_face:

mwilcox [7:55 AM]
either way, if you wait _20 minutes_ and you are mining some other block, why wouldn't you just switch pool.

csw [7:55 AM]
Then, Gun does not believe the economics benefits of risk and thinks that ONLY crypto matters.

[7:55]
:slightly_smiling_face:

mwilcox [7:58 AM]
you can't expect him to model heterogeneous networks when intel are paying the bills

[8:00]
he assumed they all have the same hashing power...

csw [8:01 AM]
Yes....

[8:01]
And the assumption is incorrect.

[8:02]
Pretty maths and a model of the wrong system.

He did not model Bitcoin.

[8:04]
If anyone is interested, I have a rather large model of the Bitcoin network, not the nodes alone, the vertices.

A complete map... The fun of years of data and a system that is said not to exist :wink:

mwilcox [8:06 AM]
would love to see that

csw [8:13 AM]
I will propose that we publish it free

[8:13]
I think it will be of value


mwilcox [8:13 AM]
no selfish data mining here!

csw [8:13 AM]
:slightly_smiling_face:

mwilcox [8:32 AM]
If you want a good laugh, read this one too. http://randomwalker.info/publications/mining_CCS.pdf

csw [8:51 AM]
I know

iang [12:16 PM]
We need a professional cite for that.  I’ve been saying that for years and people don’t get it.
csw
Bitcoin is very carefully strung together. It is to as easy as people think to just play with the system and have it work long term.
Posted in #selfish-miningToday at 7:21 AM

[12:21]
>https://btcchat.slack.com/archives/G5C87RC11/p1494763112111789
mwilcox
you can't expect him to model heterogeneous networks when intel are paying the bills
Posted in #selfish-miningToday at 7:58 AM

[12:22]
So, if he is assuming SGX which is what Intel are currently securing lots of academic work for, this assumes that CPUs are a lot more homogeneous.  I think he’s done several SGX projects if memory serves. (idle speculation, just throwing it out there)

[12:23]
Is anyone collecting these thoughts and melding them into lesson?

[12:27]
pedagogical question @csw what are the disciplines one needs to understand Bitcoin?

[12:28]
So far I see in CS: protocols, databases, state machines, stack-based languages (script)

[12:28]
in Crypto: hashes, elliptic curve

[12:29]
in Math:  distributions (power, poisson, etc), graph theory, stats including bayesian, probablility theory (edited)

iang [12:34 PM]
in Economics:  game theory, incentives, mechanism design