---
title: "Comment by Ahmet Özel on \"Part 21: Backpropagation: How Neural Networks Learn From Their Mistakes\" | Hashnode"
description: "The basketball analogy carries the intuition well. The one place it quietly breaks is worth flagging for a beginner series: the shooter knows which direction to"
url: "https://hashnode.com/posts/what-is-backpropagation/6a9b86f7c1ee45d539bb53c8/comment/6a9bf3455b6e9d2179436198"
---

[Discussion](https://hashnode.com/posts/what-is-backpropagation/6a9b86f7c1ee45d539bb53c8)

Comment [Part 21: Backpropagation: How Neural Networks Learn From Their Mistakes](https://hashnode.com/posts/what-is-backpropagation/6a9b86f7c1ee45d539bb53c8)

[Ahmet Özel](https://hashnode.com/@ahmetozel)

[Sep 5](https://hashnode.com/posts/what-is-backpropagation/6a9b86f7c1ee45d539bb53c8/comment/6a9bf3455b6e9d2179436198)

The basketball analogy carries the intuition well. The one place it quietly breaks is worth flagging for a beginner series: the shooter knows which direction to adjust because they watched where the ball landed, while a network in the middle layers has no direct feedback at all. Credit assignment is the actual hard part - working out how much each weight several layers back contributed to an error measured only at the output - and that is precisely what the chain rule is doing. Framing it that way also makes the later topics land more easily: vanishing gradients stop being a mysterious phrase and become the obvious consequence of multiplying many small numbers on the way back.

[Divyajot Kaur](https://hashnode.com/@djk2004)

[Oct 7](https://hashnode.com/posts/what-is-backpropagation/6a9b86f7c1ee45d539bb53c8/comment/6a9bf3455b6e9d2179436198#reply-6ac5ce04a2b9f7ea06509faa)

Thank you for pointing this out! 🙌 I agree that the basketball analogy leaves out an important part of the process. The real challenge is credit assignment: figuring out how much each weight, especially in the earlier layers, contributed to the final error.

I really like how you connected this to the chain rule and vanishing gradients. Thinking about it this way makes those concepts easier to understand and creates a natural link to the topics that follow. Thanks for adding this valuable insight to the discussion!

[Ahmet Özel](https://hashnode.com/@ahmetozel)

[Oct 7](https://hashnode.com/posts/what-is-backpropagation/6a9b86f7c1ee45d539bb53c8/comment/6a9bf3455b6e9d2179436198#reply-6ac65535fcba73036846cb9a)

A tiny two-layer numeric example could make that connection concrete. Compute the loss once, perturb one early weight by a small amount, and compare the observed loss change with the derivative produced by the chain rule.

Then repeat with a saturated activation. The smaller gradient becomes something a beginner can inspect, while the distinction between computing a gradient and applying the optimizer update stays visible. It would extend the analogy without suggesting that an early layer directly observes the final mistake.
